Electronic component production line optimization method using digital twinning and storage medium

Abstract

This application discloses a method and storage medium for optimizing electronic component production lines using digital twins. The method includes: mapping synchronously acquired actuator dynamic response behavior records and workstation process tolerance status records to a digital twin space; performing unsupervised behavior pattern separation processing on the actuator dynamic response behavior records to separate transient impact behavior components; performing behavioral semantic association analysis on the transient impact behavior components and workstation process tolerance status records to construct a predictive logical structure, thereby deducing the evolution trend of the process tolerance status records and the superposition effect of transient impact behavior components, generating a description of process tolerance degradation trends and a description of potential resonance workstation area distribution; and finally generating a set of production line control instructions to adjust driving force input characteristic data or action timing configuration data. In this way, the transmission impact of actuator transient impact behavior on production line process tolerance can be effectively identified and predicted.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method for optimizing electronic component production lines using digital twins and a storage medium. Background Technology

[0002] In the automated production of electronic components, production lines typically consist of multiple workstations operating in series or parallel. Each workstation outputs driving force according to preset control logic to complete the set process actions. To ensure stable operation of the production line, existing technologies generally employ periodic monitoring of the process tolerance status of each workstation. For example, sensors are used to collect tolerance status data such as vibration amplitude, temperature drift, or load fluctuation at each workstation to assess the current workstation's health. Simultaneously, for the operation control of the actuators, existing solutions focus on closed-loop feedback adjustment of the driving force input commands, that is, correcting the driving force input characteristic data for the next cycle based on the deviation of the actuator's motion output.

[0003] However, the existing technologies often treat the dynamic response behavior of the actuator and the process tolerance status of the workstation as independent monitoring objects and process them separately. They lack effective means to correlate transient dynamic behavior between workstations and affect the process tolerance status, which makes it difficult to meet the higher requirements of high-cycle electronic component production lines for the predictability and real-time control of process degradation. Summary of the Invention

[0004] This application provides a method and storage medium for optimizing electronic component production lines using digital twins. This method effectively identifies and predicts the transmission impact of transient impact behavior of actuators on the process tolerance of the production line, thereby meeting the higher requirements of high-cycle electronic component production lines for the predictability and real-time control of process degradation.

[0005] This application provides, in one aspect, a method for optimizing electronic component production lines using digital twins, applied to an electronic component production line optimization system. The method includes: Acquire the actuator dynamic response behavior record and the workstation process tolerance status record synchronously collected within a preset continuous production cycle of the target electronic component production line. The actuator dynamic response behavior record includes a description of the instantaneous transmission characteristics between the driving force input and motion output of each actuator. The dynamic response behavior record of the actuator and the process tolerance status record of the workstation are mapped to a preset digital twin space. Unsupervised behavior mode separation processing is performed on the dynamic response behavior record of the actuator in the digital twin space to separate the steady-state behavior performance component and the transient impact behavior performance component that are time-synchronized with the process tolerance status record of the workstation. Within the digital twin space, the behavioral semantic association analysis of the transient impact behavior component and the process tolerance status record of the workstation at the same time is performed to construct a predictive logical structure describing the impact of transient impact behavior on the transmission of process tolerance status. The predictive logic structure is applied to the virtual behavior of the actuator in the digital twin space to deduce the evolution trend of the process tolerance status record of the workstation in subsequent continuous production cycles and the superposition effect of the transient impact behavior component transmitted between workstations, thereby generating a description of the process tolerance decay trend and a description of the distribution of potential resonance workstation areas. Based on the description of the process tolerance degradation trend and the description of the potential resonance workstation area distribution, a set of production line control instructions is generated. The set of production line control instructions is used to adjust the driving force input characteristic data of the corresponding actuator or the action sequence configuration data between workstations.

[0006] One embodiment of this application provides an electronic component production line optimization system, including: A processor; a storage device having a computer program stored thereon; a network interface for providing network communication functions; when the computer program is executed by the processor, the processor enables the processor to implement any of the described methods for optimizing electronic component production lines using digital twins.

[0007] One embodiment of this application provides a readable storage medium storing a program or instructions, which, when executed by a processor, implements the steps of the electronic component production line optimization method using digital twins.

[0008] Therefore, the embodiments of this application have the following beneficial effects: by mapping the synchronously collected dynamic response behavior records of the actuators and the process tolerance status records of the workstations to a digital twin space, and performing unsupervised behavior mode separation processing on the dynamic response behavior records of the actuators to extract the time-synchronized steady-state behavior performance components and transient impact behavior performance components, and then performing semantic association analysis of the transient impact behavior performance components and the process tolerance status records of the workstations in the digital twin space to construct a predictive logic structure, it is possible to drive virtual actors to deduce the evolution trend of the process tolerance status records and the superposition effect of the transient impact behavior performance components between workstations based on this predictive logic structure, thereby generating a description of the process tolerance decay trend and a description of the potential resonance workstation area distribution to guide the generation of the production line control instruction set. This method realizes cross-level association from macro-production line behavior analysis to micro-dynamic response mapping, effectively improving the accuracy of process state prediction and control of electronic component production lines in continuous production cycles. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies 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.

[0010] Figure 1 This is a flowchart illustrating a method for optimizing an electronic component production line using digital twins, provided as an embodiment of this application.

[0011] Figure 2 This is a schematic diagram of the basic structure of an electronic component production line optimization system provided in an embodiment of this application.

[0012] Figure 3 This is a functional block diagram of an electronic component production line optimization device provided in an embodiment of this application. Detailed Implementation

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0014] Please see Figure 1 , Figure 1 This is a flowchart of an electronic component production line optimization method using digital twins provided in an embodiment of this application. The method can be executed by an electronic component production line optimization system, or by the electronic component production line optimization system and a server. The method may include steps 110-150.

[0015] This application provides a method for optimizing electronic component production lines using digital twins. This method analyzes and processes synchronously collected records of actuator dynamic response behavior and workstation process tolerance status during continuous production cycles. Within the digital twin space, it separates steady-state behavior components and transient impact behavior components synchronized with the process status. Based on the semantic relationship between transient impact behavior and process tolerance status, a predictive logic structure is constructed. This predictive logic structure is then used in the digital twin space to deduce the evolution trend of process tolerance and the superposition effect of impacts in subsequent production cycles. Finally, a set of production line control instructions is generated to adjust the actuator driving force input characteristics and the timing configuration of actions between workstations. This method can effectively identify and predict the transmission impact of transient impact behavior of actuators on production line process tolerance, improving the stability of production line operation and the consistency of process quality, thereby meeting the higher requirements of high-cycle electronic component production lines for predictability and real-time control of process degradation.

[0016] In this embodiment, an electronic component surface mount production line including a placement station, an insertion station, a soldering station, and an inspection station is used as the target electronic component production line for description. This production line is equipped with multiple actuators, such as the XY-axis motion platform of the placement machine, the Z-axis lifting and rotation mechanism of the placement head, the gripper opening and closing and insertion mechanism of the insertion machine, and the stepper drive mechanism of the conveyor belt. The digital twin space of the production line has been pre-constructed and synchronized with the physical production line at millisecond levels via an industrial IoT gateway. This digital twin space contains virtual behavior simulation units corresponding one-to-one with each physical actuator. Each virtual behavior simulation unit is equipped with a dynamic model describing its motion transmission attributes, capable of mapping the motion state of the physical actuator in real time.

[0017] Step 110: Obtain the actuator dynamic response behavior record and the workstation process tolerance status record synchronously collected within the preset continuous production cycle of the target electronic component production line.

[0018] In this step, the preset continuous production cycle is set as the continuous operation period of the electronic component production line from the start of a shift to the current moment. The actuator dynamic response behavior record contains a description of the instantaneous transmission characteristics between the driving force input and motion output of each actuator.

[0019] For the X-axis motion platform of the pick-and-place machine, the driving force input is the current loop command sequence of the servo motor, and the motion output is the actual position sequence fed back by the grating ruler. The instantaneous transmission characteristic is described as the amplitude-frequency and phase-frequency relationship between the current command and the actual acceleration. Specifically, it is the correspondence between the rising edge slope of the driving current waveform and the acceleration settling time of the motion platform.

[0020] For the Z-axis lifting mechanism of the placement head, the instantaneous transmission characteristics are described by the drift curve of the voice coil motor thrust constant as a function of coil temperature and the micron-level jitter amplitude of the moving parts at the moment of start-up and stop. The station process tolerance status record refers to a set of records that quantitatively describe the process stress and process deviations experienced by key process stations on the production line within the same continuous production cycle.

[0021] At the placement station, the process tolerance record specifically includes the contact force waveform curve collected by the placement force sensor during the placement head's downward pressure. This curve includes the peak impact force at the moment of placement contact, the force fluctuation range during the steady-state holding pressure phase, and the abrupt change in release force when the placement head is lifted. At the soldering station, the process tolerance record includes the hot air flow rate fluctuation sequence in each temperature zone of the reflow oven cavity, as well as the component pin offset sequence extracted by automatic optical inspection after soldering.

[0022] The acquisition process of the actuator dynamic response behavior record and the workstation process tolerance status record is kept in time synchronization. The synchronization accuracy is guaranteed by the precision time protocol of industrial Ethernet, and the deviation is controlled at the microsecond level, so that each dynamic response behavior record can be accurately matched with the process tolerance status record at the same moment.

[0023] Step 120: Map the actuator dynamic response behavior record and the workstation process tolerance state record to a preset digital twin space. Perform unsupervised behavior mode separation processing on the actuator dynamic response behavior record in the digital twin space to separate the steady-state behavior performance component and transient impact behavior performance component that are time-synchronized with the workstation process tolerance state record.

[0024] The electronic component production line optimization system injects the raw record data obtained in step 110 into the digital twin space through a data interface. Within the digital twin space, based on the timestamp of the data packet and the workstation identifier, the dynamic response behavior records of the actuators are first written into the state observer buffer of the corresponding virtual behavior simulation unit, and the process tolerance status records of the workstation are written into the process status description module of the corresponding virtual workstation.

[0025] At this point, the virtual placement machine, virtual insertion machine, and other actors in the digital twin space are all loaded with behavioral state data that is completely consistent with the physical production line at the same moment. In order to decouple the components reflecting long-term stable operation from the components reflecting transient process disturbances from the coupled dynamic response behavior, this step performs unsupervised behavior mode separation processing.

[0026] Step 121: Analyze the instantaneous transmission characteristic description of each actuator in the actuator dynamic response behavior record, and extract the continuous change profile curve of the power transmission gain change over time and the phase drift trajectory curve of the cumulative value of hysteresis phase offset over time.

[0027] For the X-axis motion platform of a pick-and-place machine, in the description of its instantaneous transmission characteristics, the change in power transmission gain is defined as the change in the ratio of the actual acceleration response amplitude to the drive current command amplitude over continuous operation time. During extraction, the acquired drive current command sequence and acceleration response sequence are first divided into several analysis segments of equal duration according to a time window. Within each analysis segment, the cross-power spectral density of current and acceleration is calculated, and the gain amplitude at the corresponding dominant motion frequency point is extracted. The gain amplitudes extracted from each analysis segment are connected in chronological order to form a continuous profile curve of the evolution of the power transmission gain over time.

[0028] The cumulative hysteresis phase shift refers to the phase lag of the actual position response relative to the commanded position. The extraction process involves calculating the time delay corresponding to the peak value of the cross-correlation function between the commanded position sequence and the feedback position sequence within each analysis segment, converting this time delay into a phase angle, and accumulating the phase lag values ​​calculated from all analysis segments along the time axis to form a phase drift trajectory curve of the cumulative hysteresis phase shift over time.

[0029] Step 122: Establish virtual behavior simulation units corresponding to each actuator in the physical production line in the preset digital twin space, and load the continuously changing contour curve and the phase drift trajectory curve into the motion transmission attribute description model of the corresponding virtual behavior simulation unit to generate a virtual mechanism dynamic behavior representation that is synchronized with the behavior of the physical actuator in real time.

[0030] The digital twin space already contains the geometric models and kinematic constraints of each actuator. The motion transmission attribute description model is a mathematical model describing the dynamic relationship between the driving force input and motion output. For the X-axis platform of the pick-and-place machine, its motion transmission attribute description model is a second-order mass-spring-damped system model with three adjustable attributes: mass parameter, damping coefficient, and stiffness coefficient. The continuously changing profile curve extracted in step 121 is used to dynamically correct the damping coefficient in this model, so that its change over time reflects the changing trend of the lubrication state of the mechanical guide rail. The phase drift trajectory curve is used to dynamically correct the stiffness coefficient in the model, so that it reflects the creep effect of the drive belt or coupling under long-term operation.

[0031] After loading the corrected model parameters, the virtual behavior simulation unit can output a multi-dimensional behavior representation vector, which includes time series such as virtual position, virtual velocity, virtual acceleration, and virtual motor torque. The set of this multi-dimensional behavior representation vector is the virtual mechanism dynamic behavior representation body. The behavior of this virtual mechanism dynamic behavior representation body is highly synchronized with the behavior of the physical actuator under the same driving force input.

[0032] Step 123: Perform an unsupervised blind separation operation based on self-similarity metric on the behavior representation signal sequence output by the virtual mechanism dynamic behavior representation body to decompose the behavior representation signal sequence into several behavior pattern components with different fluctuation characteristics.

[0033] This step employs the variational mode decomposition algorithm as the specific implementation of the unsupervised blind separation operation of behavioral modes. The time series representing virtual acceleration in the virtual mechanism's dynamic behavior representation is input into the variational mode decomposition algorithm as the behavior representation signal sequence. The algorithm iteratively searches to determine the optimal number of mode decompositions K, as well as the center frequency and finite bandwidth of each mode component.

[0034] During the iteration process, the algorithm constructs and solves a variational optimization problem with the objective function of minimizing the sum of the estimated bandwidths of each modal component. The constraint is that the sum of all modal components equals the original behavioral representation signal sequence. This optimization problem is solved using the alternating direction multiplier method, ultimately decomposing the original virtual acceleration time series into K intrinsic mode function (IMF) components. Each IMF component represents a fluctuation component at a set frequency scale in the original behavioral representation signal, with different fluctuation frequency ranges and amplitude variation patterns. For example, low-frequency components correspond to the slow preheating expansion of the actuator or the guide rail wear process, while high-frequency components correspond to the impact oscillations at the start and stop of the mechanism or micro-collisions caused by transmission chain gaps.

[0035] Step 124: Calculate the mutual information correlation strength between each of the behavioral pattern components and the workstation assembly contact force waveform curve in the workstation process tolerance state record.

[0036] The contact force waveform curves at the assembly station are taken from the mounting force sensor records at the placement station. First, the contact force waveform curves are downsampled to the same time resolution as the behavioral mode components. Then, a mutual information estimation algorithm based on K-nearest neighbor distance is used to calculate the mutual information value between each intrinsic mode function component and the contact force waveform curve.

[0037] The mutual information value is calculated based on the following logic: the edge probability density of a single intrinsic mode function component sequence, the edge probability density of a contact force waveform sequence, and the joint probability density of the two are estimated separately. The mutual information value is then obtained by calculating the difference between the sum of the edge entropies and the joint entropy. A larger mutual information value indicates a stronger statistical dependence between the fluctuation of the behavioral mode component and the fluctuation of the process contact force; that is, the more likely the behavioral mode component is to carry dynamic information directly related to process impact events.

[0038] Step 125: Classify behavioral pattern components with mutual information correlation strength lower than the first preset judgment level as steady-state pattern component candidate set, and classify behavioral pattern components with mutual information correlation strength higher than the second preset judgment level as transient impact pattern component candidate set.

[0039] The first and second preset judgment levels are mutual information thresholds predetermined based on statistical analysis of historical operating data, denoted as threshold Mth1 and threshold Mth2 respectively, with Mth2 being greater than Mth1. For the mutual information value MIk of each intrinsic mode function component calculated in step 124, the following judgment is performed: if MIk is less than Mth1, the component is assigned to the steady-state mode component candidate set; if MIk is greater than Mth2, the component is assigned to the transient impact mode component candidate set; if MIk is between Mth1 and Mth2, it is considered an ambiguous component and is not assigned to any candidate set for the time being.

[0040] Step 126: By filtering the candidate set of steady-state mode components and the candidate set of transient impact mode components, the steady-state behavior performance components and the transient impact behavior performance components are obtained and time-synchronized.

[0041] To extract the most representative behavioral components from the candidate set, further filtering and merging operations are required.

[0042] Step 1261: Cluster and merge the behavioral pattern components in the steady-state mode component candidate set according to their fluctuation frequency range to obtain the steady-state behavioral performance components that characterize the long-term stable operating state of the actuator.

[0043] First, the center frequency of each intrinsic mode function component in the steady-state mode component candidate set is extracted. A hierarchical clustering algorithm is then used, with the Euclidean distance between center frequencies as a similarity measure, to aggregate components with similar frequencies into several clusters. For each aggregated cluster, all intrinsic mode function components contained within it are linearly superimposed to generate a merged mode component.

[0044] Then, from the merged modal components, the modal component with the lowest center frequency and whose amplitude exhibits a monotonically slow change trend over time is selected as the steady-state behavior component. This steady-state behavior component includes the baseline drift trend parameter of the power transfer gain and the low-frequency cumulative change characteristic of the phase drift. The baseline drift trend parameter is represented by the slope of the amplitude envelope of this component, and the low-frequency cumulative change characteristic is represented by the integral value of the instantaneous frequency of this component over a long time scale.

[0045] Step 1262: Select each behavioral mode component in the transient impact mode component candidate set according to the time overlap of its peak transient amplitude parameter with the waveform curve of the assembly contact force at the workstation, and obtain the transient impact behavior performance component that characterizes the instantaneous change of the mechanism behavior caused by the process impact event.

[0046] The peak transient amplitude parameter of the contact force waveform curve at the workstation refers to the amplitude change corresponding to each peak exceeding a preset amplitude threshold in the waveform. Time coincidence is defined as whether the time difference between the amplitude jump moment in the behavior mode component and the peak transient moment of the contact force waveform is less than a preset time window.

[0047] For each component in the candidate set of transient impact mode components, the time difference between each jump moment and the most recent peak contact force transient moment is calculated. If a component statistically satisfies the condition that the time difference between a jump event exceeding a preset proportional threshold and the peak contact force event is less than a preset time window, then the component is identified as a transient impact behavior component. This transient impact behavior component includes the impact response amplitude decay envelope and the phase reset dynamic process characteristics after the impact disturbance. The impact response amplitude decay envelope is represented by a contour line showing an exponential decay of the component amplitude over time after the jump, and the phase reset dynamic process characteristics are represented by the trajectory of the component's instantaneous phase undergoing a brief nonlinear adjustment process after the impact disturbance and then stabilizing at a certain phase value.

[0048] Step 1263: The steady-state behavior component and the transient impact behavior component are respectively calibrated with the corresponding workstation process tolerance state record on the time axis, so that the steady-state behavior component and the transient impact behavior component are synchronized with the workstation process tolerance state record in time.

[0049] Since the mode decomposition and screening process may introduce a small group delay, the separated components need to be aligned on the time axis. Using the timestamp sequence of the process tolerance status record at the workstation as a reference, the time series of the steady-state behavior performance components and the transient impact behavior performance components are resampled and phase-corrected. The correction amount is determined by calculating the peak position of the cross-correlation function between the component and the reference sequence.

[0050] After unified calibration, the three time series—steady-state behavior component series, transient impact behavior component series, and workstation process tolerance state record series—correspond one-to-one at each time point.

[0051] Step 130: In the digital twin space, perform behavioral semantic association analysis on the transient impact behavior performance component and the process tolerance state record of the workstation at the same time to construct a predictive logical structure describing the impact of transient impact behavior on the transmission of process tolerance state.

[0052] This step aims to uncover the intrinsic causal or correlational relationship between the impact characteristics in the transient impact behavior component and the process deviation characteristics in the process tolerance status record of the workstation, and to solidify this relationship into a logical structure that can be used for prediction.

[0053] Step 131: Extract the initial amplitude parameter of the impact response amplitude attenuation envelope and the recovery time length parameter of the phase reset dynamic process characteristics after impact disturbance from the transient impact behavior performance components.

[0054] In the transient impact behavior component, each jump event is considered a process impact event. For the identified q-th process impact event, the envelope amplitude of the transient impact behavior component at the time of the event is extracted as the initial amplitude parameter Ainitq. Simultaneously, the time taken for the instantaneous phase of the transient impact behavior component to recover from the disturbed unsteady state to the steady-state phase neighborhood after the event is recorded, and this time is used as the recovery time length parameter Trecq.

[0055] Step 132: Extract the peak transient amplitude parameter and the time-domain fluctuation range parameter of the assembly contact force waveform curve synchronized with the time of each process impact event from the process tolerance status record of the workstation, and generate a process tolerance instantaneous deviation record pair.

[0056] Synchronously with the q-th process impact event, the difference between the maximum contact force amplitude near the corresponding moment of the event and the steady-state contact force amplitude before the event is extracted from the assembly contact force waveform curve recorded by the mounting force sensor, and used as the peak transient amplitude parameter Fpeakq. Simultaneously, the standard deviation of the deviation between the actual component placement position and the target position within the first placement cycle after the event is extracted from the component placement position feedback data recorded by the pick-and-place machine vision system, and used as the time-domain fluctuation range parameter Dstdq of the assembly positioning feedback deviation value. The parameter pair (Fpeakq, Dstdq) is used as the instantaneous deviation record pair of process tolerance corresponding to the q-th impact event.

[0057] Step 133: Establish a process shock risk correlation model in the digital twin space. The process shock risk correlation model uses the initial amplitude parameter and the recovery time length parameter as the set of input variables, and the peak transient amplitude parameter and the time domain fluctuation range parameter as the set of output observation variables.

[0058] The process shock risk correlation model is implemented using a dual-input, dual-output adaptive neural fuzzy inference system. The input layer has two nodes, corresponding to the initial amplitude parameter Ainit and the recovery time length parameter Trec, respectively. The output layer has two nodes, corresponding to the peak transient amplitude parameter Fpeak and the time-domain fluctuation range parameter Dstd, respectively. Internally, the system includes a fuzzification layer, a rule layer, a normalization layer, and a defuzzification layer.

[0059] The fuzzification layer defines several membership functions for each input variable. For example, Ainit is fuzzified into three fuzzy sets: "minor impact", "moderate impact", and "strong impact", and Trec is fuzzified into three fuzzy sets: "rapid recovery", "normal recovery", and "slow recovery". The rule layer contains a set of fuzzy rules preset based on prior process knowledge, such as "if the impact is strong and the recovery is slow, the peak contact force will be high and the positioning deviation will fluctuate drastically".

[0060] Step 134: Learn the co-evolutionary transitive relationship between the set of input variables and the set of output observation variables through unsupervised association rule mining, and obtain the transfer function description of the input variables to the output observation variables.

[0061] A fuzzy C-means clustering algorithm based on subtractive clustering is used to perform cluster analysis on all four-dimensional data samples (Ainitq, Trecq, Fpeakq, Dstdq) generated in steps 131 and 132. The subtractive clustering algorithm automatically determines the number and location of cluster centers. Each cluster center is used as an initial parameter for a fuzzy rule to construct an initial Sugeno-type fuzzy inference system.

[0062] Then, a hybrid learning algorithm is used to optimize and adjust the antecedent parameters (parameters of the membership function) and consequent parameters (linear coefficients of the rule output function) of the fuzzy inference system. In the forward propagation phase, the hybrid learning algorithm fixes the antecedent parameters and uses a least-squares estimator to identify the consequent parameters; in the backward propagation phase, it fixes the consequent parameters and uses gradient descent to optimize the antecedent parameters. This process is iterated until the preset convergence accuracy is reached. After training, the fuzzy inference system can map any input combination (Ainit, Trec) to an output combination (Fpeak, Dstd), and this mapping relationship is described by the transfer function of the input variables to the output observed variables.

[0063] Step 135: Based on the process tolerance instantaneous deviation record pairs corresponding to multiple sets of process shock events occurring sequentially during the continuous operation period, calculate the offset direction and offset rate of the gain coefficient described by the transfer function as the continuous operation time progresses.

[0064] The continuous operation period is divided into multiple equal-length evaluation cycles in chronological order, for example, one evaluation cycle is every 1000 production cycles. Within each evaluation cycle, the fuzzy inference system in step 134 is retrained or incrementally updated using data samples of all impact events that occurred within the cycle to obtain the transfer function description corresponding to that cycle.

[0065] Then, the output change of the transfer function under the same input conditions is calculated for each evaluation period. A set of benchmark inputs is selected (e.g., historical averages for Ainit and Trec), and the transfer function for each evaluation period is input respectively to obtain the corresponding outputs Fpeakest and Dstdest. The ratio of the difference in output values ​​between adjacent evaluation periods to the time interval is calculated to obtain the offset rate of the gain coefficient described by the transfer function in the two dimensions of Fpeak and Dstd, which are denoted as the offset rate vector Vdrift. The offset direction is represented by the direction angle of Vdrift in each evaluation period.

[0066] Step 136: Based on the offset direction and offset rate of the gain coefficient, construct a predictive logic structure to describe the impact of transient shock behavior on the transmission of process tolerance status.

[0067] A predictive logical structure is a multi-level set of logical relationships.

[0068] First, the predictive logic structure includes a monotonic mapping relationship between the gain coefficient offset rate and the trend of the tolerance state deviation expansion. This mapping relationship indicates that when the magnitude of the offset rate vector Vdrift exceeds the preset rate threshold Spdth, the trend of the process tolerance state deviation expansion will exhibit accelerated degradation characteristics, and the degree of acceleration is positively correlated with the proportion of the offset rate magnitude exceeding the threshold.

[0069] Secondly, the predictive logic structure includes a correlation description between the gain coefficient offset direction and the tendency of potential resonance risk in subsequent workstations. This correlation description indicates that if the direction angle of the offset rate vector points to the set spatial quadrant corresponding to the subsequent workstation, it indicates that the cumulative effect of impact energy on the transmission path tends to induce mechanical resonance or process parameter coupling oscillation in that subsequent workstation.

[0070] Furthermore, the predictive logic structure also includes a description of the propagation attenuation characteristics of transient impact behavior components as they are transmitted between multiple workstations on the production line. This propagation attenuation characteristic description defines the attenuation coefficient sequence attseq of the initial amplitude parameter when the impact response is transmitted from workstation i to the adjacent workstation i+1. Each element attcoefk in the attenuation coefficient sequence represents the proportion of the remaining amplitude after the impact passes through the k-th structural connection link. This attenuation coefficient sequence is determined based on the correlation analysis results between the transient impact behavior components corresponding to adjacent workstations. Specifically, it is determined by calculating the Pearson correlation coefficient between the envelope signals of the transient impact behavior components of workstation i and workstation i+1. The higher the correlation coefficient, the smaller the attenuation coefficient, indicating lower transmission loss.

[0071] Step 140: Apply the predictive logic structure to the virtual behavior of the actuator in the digital twin space, deduce the evolution trend of the process tolerance status record of the workstation in subsequent continuous production cycles and the superposition effect of the transient impact behavior component transmitted between workstations, and generate a description of the process tolerance decay trend and a description of the potential resonance workstation area distribution.

[0072] After completing the construction of the predictive logic structure, the electronic component production line optimization system initiates a real-time forward simulation process based on a digital twin space. The execution of this step relies on a virtual production line simulation environment that runs independently in the digital twin space. This environment accurately replicates the current physical production line state, and its time progression is much faster than the flow of time in the physical world.

[0073] The simulation process integrates a closed-loop iterative calculation flow that combines the mechanism dynamics model, process state transition rules, and inter-station coupling and transmission mechanism. Within each simulation step, the state of each virtual behavior simulation unit is updated and the transmission and superposition effects of the impact between stations are calculated.

[0074] Step 141: Load the predictive logic structure into the virtual production line simulation environment in the digital twin space.

[0075] The initialization of the virtual production line simulation environment requires converting the previously obtained logical rules and transfer functions into a computable rule set. An independent simulation region is partitioned within the memory space of the digital twin space. The state parameters of each virtual behavior simulation unit at the current moment are completely copied to form an initial state snapshot. Then, each descriptive rule in the predictive logic structure is parsed and converted into a mathematically executable update function.

[0076] Step 1411: Instantiate a virtual production line simulation environment in the digital twin space and inherit the snapshot of the state parameters of each virtual behavior simulation unit at the current moment from the digital twin space.

[0077] When a simulation task is triggered, the system calls the environment instantiation interface to create an independent simulation container in the parallel computing resource pool of the digital twin space. This container performs a deep copy operation from the state parameter buffer of the main thread. The inherited state parameter snapshot content includes the current instantaneous amplitude parameter and the current instantaneous phase parameter of the steady-state behavior performance component corresponding to each virtual behavior simulation unit, as well as the current remaining amplitude parameter of the impact response amplitude decay envelope of the transient impact behavior performance component corresponding to the most recent process impact event experienced by the unit and the current recovery stage identifier of the phase reset dynamic process.

[0078] For the virtual behavior simulation unit of the X-axis motion platform of the pick-and-place machine, the instantaneous amplitude parameter of the steady-state behavior performance component reflects the current quantization level of its guideway damping characteristics. This parameter is determined by the sampled value of the amplitude envelope of the steady-state behavior performance component extracted in step 1261 at the current moment. The instantaneous phase parameter reflects the current cumulative phase lag in its position tracking. The current residual amplitude parameter of the impact response amplitude decay envelope reflects the magnitude of the residual vibration caused by the previous placement impact that has not yet fully decayed at the current moment. This parameter is calculated by recursively extrapolating the initial amplitude parameter of the previous impact event to the current moment according to the exponential decay law. The recovery phase identifier indicates whether the phase reset dynamic process is currently in the disturbed period, recovery period, or steady-state period. All inherited parameters are timestamped with a time stamp synchronized with the physical clock, forming the initial time starting point of the simulation.

[0079] Step 1412: The monotonic mapping relationship between the gain coefficient offset rate and the tolerance state deviation expansion trend in the predictive logic structure is described and converted into a state transition rule for updating the process tolerance index of each workstation with the simulation step size in the virtual production line simulation environment.

[0080] Each simulation step corresponds to the time interval required for the physical production line to complete one full production cycle. The state transition rule transformation converts the description of the monotonic mapping relationship between the gain coefficient offset rate and the trend of the increase in tolerance state deviation established in step 136 into a numerical update function that can be directly called by the simulation engine. First, the preset rate threshold and accelerated degradation feature defined in the monotonic mapping relationship description are parameterized into the threshold parameter and acceleration factor calculation function in the state transition rule.

[0081] At the start of each simulation step, the state transition rule obtains the magnitude of the offset rate vector of the transient impact behavior component corresponding to each actuator under the current step. The value of this magnitude is calculated by the offset rate calculation method described in step 135 within the evaluation period corresponding to the current simulation step, and is dynamically updated during the simulation process as the state of the virtual behavior simulation unit evolves. Then, the current offset rate magnitude is compared with a preset rate threshold. If the offset rate magnitude is less than or equal to the preset rate threshold, the degradation acceleration factor is taken as the baseline value. If the offset rate magnitude is greater than the preset rate threshold, the degradation acceleration factor is taken as the baseline value plus the product of the excess and a preset slope coefficient. This preset slope coefficient is determined based on the statistical regression coefficient between the offset rate magnitude and the degree of degradation acceleration obtained from the historical data analysis in step 135.

[0082] Finally, the process tolerance index for each workstation is updated as follows: the process tolerance index value for the current simulation step is equal to the process tolerance index value for the previous simulation step multiplied by the degradation acceleration factor. For the patch workstation, its process tolerance index specifically refers to the time-domain fluctuation range parameter of the assembly positioning feedback deviation value. This parameter is updated and stored at the end of each simulation step, forming the predicted expansion trajectory of this parameter over time.

[0083] Step 1413: The correlation description between the gain coefficient offset direction in the predictive logic structure and the tendency of subsequent workstations to trigger potential resonance risks is converted into a coupling strength adjustment rule for the transmission of dynamic behavior between adjacent workstations in the virtual production line simulation environment.

[0084] In the topology model of the virtual production line simulation environment, a reference coupling strength coefficient is defined for the connection between each pair of adjacent workstations. This reference coupling strength coefficient is preset based on the physical connection method and structural stiffness between adjacent workstations. The chip mounting workstation and the component insertion workstation are connected by a rigid substrate transport track, and their reference coupling strength coefficient is set to a higher value. The component insertion workstation and the soldering workstation are connected by a flexible conveyor belt, and their reference coupling strength coefficient is set to a lower value. Within each simulation step, the cosine of the angle between the gain coefficient offset direction vector at the current step and the spatial direction vector pointing from the current workstation to the subsequent workstation is calculated. The value of this cosine angle ranges from -1 to +1. The closer the value is to +1, the more accurately the offset direction points to the subsequent workstation.

[0085] The coupling strength adjustment rule defines the effective coupling strength coefficient as the product of a baseline coupling strength coefficient and an enhancement factor. The enhancement factor is calculated as follows: when the cosine of the included angle is less than or equal to zero, the enhancement factor is the baseline value; when the cosine of the included angle is greater than zero, the enhancement factor is the baseline value plus the product of the cosine of the included angle and a preset enhancement amplitude coefficient. The preset enhancement amplitude coefficient is determined based on offline simulation results and is used to control the increase in coupling strength during orientation alignment. This coupling strength adjustment rule is invoked within each simulation step to dynamically adjust the interaction stiffness between virtual behavior simulation units of adjacent workstations.

[0086] Step 1414: The propagation attenuation characteristics of the transient impact behavior component in the predictive logic structure as it propagates through multiple workstations on the production line are transformed into the amplitude attenuation coefficient sequence and phase delay accumulation rule when the impact response is propagated between multi-level virtual behavior simulation units in the virtual production line simulation environment.

[0087] Each attenuation coefficient in the attenuation coefficient sequence is bound to a corresponding structural connection link in the virtual production line topology. Structural connection links include connections between kinematic pairs within the actuator, connections between the actuator and the workstation base, and connections between the workstation base and the substrate transport track. The amplitude attenuation calculation rule is defined as follows: when an impact behavior characterization signal is transmitted from the i-th virtual behavior simulation unit to the (i+1)-th virtual behavior simulation unit, its amplitude is updated after passing through the k-th structural connection link to the amplitude before passing through that link multiplied by the attenuation coefficient corresponding to that link. If the transmission path contains multiple structural connection links, the amplitude attenuation effect is the product of the attenuation coefficients of each link. The phase delay accumulation rule stipulates that for each structural connection link the impact behavior characterization signal passes through, its phase delay increases by a preset base phase delay value. This base phase delay value is related to the physical characteristics of the structural connection link and is determined by its equivalent stiffness and equivalent mass.

[0088] During the simulation, as the impact signal propagates through multiple stages between workstations, its amplitude gradually attenuates while its phase gradually lags behind. The combined effect of the amplitude attenuation coefficient sequence and the phase delay accumulation rule makes the simulated impact transmission process in the simulation environment physically plausible.

[0089] Step 1415: In the virtual production line simulation environment, start the simulation clock and execute the state transition rule, the coupling strength adjustment rule, and the amplitude attenuation coefficient sequence and phase delay accumulation rule in a cyclical manner according to the preset simulation step size, so as to drive the behavior state of each virtual behavior simulation unit to evolve forward in sequence.

[0090] After completing initialization and rule transformation, the system starts the simulation clock in the virtual production line simulation environment. The initial time of the simulation clock is set to the simulation time zero point, which corresponds to the time of inheriting the state parameter snapshot in step 1411. The simulation clock jumps forward in preset simulation step sizes, and executes various rules in a fixed order within each simulation step size. First, the state transition rules are executed to update the value of the process tolerance index of each station in the current simulation step size, and the process tolerance decay trajectory is recorded.

[0091] For each virtual behavior simulation unit, its motion response within the simulation step is calculated based on its current driving force input setting and motion transmission attribute description model, and the transient impact behavior component generated within that step is extracted. The extraction of the transient impact behavior component is achieved by detecting amplitude jump events in the virtual acceleration signal and extracting their amplitude attenuation envelope and phase reset process features. Then, the amplitude attenuation coefficient sequence and phase delay accumulation rule are executed to transmit the transient impact behavior component generated by each virtual behavior simulation unit to adjacent downstream units according to the production line topology, with corresponding attenuation and delay processing applied to its amplitude and phase during the transmission process.

[0092] Then, the coupling strength adjustment rule is executed, dynamically adjusting the interaction coupling strength between adjacent units according to the gain coefficient offset direction of the current step size, affecting the efficiency of impact transmission in subsequent steps. After completing the state update of all virtual behavior simulation units, the simulation clock advances one step, and the system enters the next simulation cycle until the preset total simulation time is reached.

[0093] Step 142: Based on the monotonic mapping relationship between the gain coefficient offset rate and the tolerance state deviation expansion trend in the predictive logic structure, the transient impact behavior performance components of each actuator in the virtual production line simulation environment are predicted to evolve within a preset continuous operation period, thereby obtaining the predicted initial amplitude sequence and predicted recovery time sequence of the transient impact behavior performance components within the preset continuous operation period.

[0094] During the cyclical simulation in the virtual production line simulation environment, the transient impact behavior components of each actuator at future moments need to be predicted within each simulation step. The prediction is based on the monotonic mapping relationship established in step 136, which implicitly describes the evolution of the offset rate itself with time and the state of the actuator. Within each simulation step, the initial amplitude parameters and recovery time parameters of the transient impact behavior components within the current simulation step are predicted using the state information from the previous simulation step. The predicted sequence of these two parameters will serve as the input to the transfer function description in step 143.

[0095] Step 1421: Within each simulation step of the virtual production line simulation environment, obtain the initial amplitude parameters and recovery time length parameters of the transient impact behavior performance components corresponding to each actuator in the previous simulation step.

[0096] At the start of the p-th simulation step, the system reads the transient impact behavior performance component status record stored at the end of the p-1st simulation step. This status record contains a complete description of the process impact events experienced by each actuator within the p-1st step. For the X-axis motion platform of the pick-and-place machine, the initial amplitude parameters and recovery time parameters of the impact events occurring within this step are recorded.

[0097] The initial amplitude parameter characterizes the instantaneous amplitude of the transient impact behavior component envelope at the moment of the jump, reflecting the severity of the impact. The recovery time parameter characterizes the time span during which the instantaneous phase of the transient impact behavior component recovers from the disturbed state to the steady-state phase neighborhood after the impact event, reflecting the ability of the actuator control system to suppress the impact disturbance. If no impact event occurs within the p-th minus one iteration, the system uses the parameters of the most recent impact event as the reference value for the current state.

[0098] Step 1422: Based on the monotonic mapping relationship between the gain coefficient offset rate and the expansion trend of the tolerance state deviation, determine the incremental adjustment direction and adjustment magnitude of the initial amplitude parameter of the transient impact behavior performance component within the current simulation step size relative to the previous simulation step size.

[0099] When the deviation in process tolerance shows an increasing trend, it means that the power transmission characteristics of the actuator are deteriorating. This deterioration, in turn, will cause the transient impact behavior generated during the movement of the mechanism to become more severe under the same driving force input conditions. First, obtain the offset rate vector calculated in the previous simulation step and extract its magnitude and direction. According to the monotonic mapping relationship, the larger the offset rate magnitude, the faster the system is currently in a deterioration channel. Correspondingly, the initial amplitude parameter of the newly occurring impact event within the current simulation step should also show an increasing trend. The incremental adjustment direction is in the same direction as the gain coefficient offset direction. When the offset direction is positive, the incremental adjustment direction is also positive, and when the offset direction is negative, the incremental adjustment direction is also negative. The adjustment magnitude is proportional to the offset rate magnitude. The increment value is equal to the offset rate magnitude multiplied by the simulation step time and then multiplied by the amplitude change sensitivity coefficient. The amplitude change sensitivity coefficient is determined based on the linear regression slope between the impact amplitude change and the offset rate magnitude in historical data, and its dimensions ensure that the units on both sides of the equation remain consistent.

[0100] Step 1423: Update the predicted initial amplitude parameters of the transient impact behavior manifestation components within the current simulation step size according to the incremental adjustment direction and adjustment magnitude, and apply the updated predicted initial amplitude parameters to the impact response amplitude attenuation envelope generation process of the virtual behavior simulation unit within the current simulation step size.

[0101] The predicted initial amplitude parameter is equal to the initial amplitude parameter of the previous simulation step plus the increment value. The updated predicted initial amplitude parameter is assigned to the shock response amplitude decay envelope generation function of the virtual behavior simulation unit within the current simulation step as the peak parameter of that envelope. In the subsequent dynamic solution of the virtual behavior simulation unit, when the simulated shock event is triggered, the system will use this predicted initial amplitude parameter as the initial amplitude of the shock response amplitude decay envelope, and generate an exponentially decaying envelope curve according to a preset decay time constant, superimposed on the behavioral characterization signal of the virtual behavior simulation unit. This process ensures that the transient shock behavior simulated in the simulation environment is consistent with the evolution trend predicted by the predictive logic structure in terms of amplitude characteristics.

[0102] Step 1424: Based on the relative relationship between the updated predicted initial amplitude parameter and the baseline drift trend parameter of the power transmission gain in the steady-state behavior component, adjust the predicted recovery time length parameter of the transient impact behavior component within the current simulation step size, so that the predicted recovery time length parameter monotonically increases as the predicted initial amplitude parameter increases.

[0103] Predicting the recovery time parameter requires comprehensive consideration of both the severity of the shock and the system's current damping recovery capability. The more severe the shock, the greater the disturbance to the system, and the longer the adjustment time required to recover to steady state.

[0104] As continuous operation time increases, the baseline drift trend parameter of the actuator's power transmission gain typically shows a downward trend, reflecting the deterioration of the mechanism's damping characteristics. Reduced damping also leads to slower vibration decay after impact. First, the ratio between the updated predicted initial amplitude parameter and the current power transmission gain baseline drift trend parameter is calculated. This ratio comprehensively reflects the relative relationship between impact intensity and system damping capability. The predicted recovery time parameter is multiplied by an adjustment factor positively correlated with this ratio, based on the recovery time parameter of the previous simulation step. The adjustment factor equals the baseline value plus the product of the ratio and a preset proportional coefficient. This adjustment mechanism allows the predicted recovery time parameter to adaptively and dynamically adjust to changes in impact amplitude and system damping.

[0105] Step 1425: Store the predicted initial amplitude parameter and the predicted recovery time length parameter generated at each simulation step in the order of simulation step to generate the predicted initial amplitude sequence and the predicted recovery time length sequence of the transient impact behavior performance components within the preset continuous operation period.

[0106] After each simulation step is calculated, the system writes the predicted initial amplitude parameter and the predicted recovery time length parameter generated by that step, along with the simulation step number and timestamp, into the designated storage area of ​​the simulation log.

[0107] As the simulation clock advances, the accumulated records in the storage area gradually form a predicted initial amplitude sequence and a predicted recovery time sequence. The predicted initial amplitude sequence reflects the expected evolution trajectory of the impact severity of each process impact event occurring on the X-axis motion platform of the pick-and-place machine within a preset continuous operation period. Under typical degradation scenarios, this sequence exhibits a monotonically increasing trend. The predicted recovery time sequence reflects the expected evolution trajectory of the time required for the system to recover to steady state after each impact event. Under typical degradation scenarios, this sequence also exhibits a monotonically increasing trend. These two sequences provide a quantitative basis for subsequent analysis of the health degradation patterns of the actuators.

[0108] Step 143: Input the predicted initial amplitude sequence and the predicted recovery duration sequence into the transfer function description in the predictive logic structure to deduce the predicted value sequence of the peak transient amplitude parameter of the assembly contact force waveform curve of each workstation within the preset continuous operation period and the predicted expansion trajectory of the time domain fluctuation range parameter of the assembly positioning feedback deviation value.

[0109] For each inference step, the system constructs a two-dimensional input vector from the predicted initial amplitude parameter and the predicted recovery time parameter obtained at that step, and feeds it into the adaptive neural fuzzy inference system trained in step 134. The input vector first passes through a fuzzification layer, where the membership degree of the input vector to each fuzzy set is calculated by various membership functions. The rule layer calculates the excitation intensity of each rule based on a pre-defined fuzzy rule library. The normalization layer normalizes the excitation intensity of each rule to form rule weights. The defuzzification layer uses a weighted average method to synthesize the outputs of all rules and calculates the predicted peak transient amplitude parameter and the predicted time-domain fluctuation range parameter corresponding to that step.

[0110] The system performs the above inference calculations for each step size, and concatenates the outputs of all steps sequentially to obtain the predicted value sequence of the peak transient amplitude parameter and the predicted expansion trajectory of the time-domain fluctuation range parameter. The predicted expansion trajectory is represented by a curve that rises continuously or accelerates over time, with the vertical axis representing the predicted value of the time-domain fluctuation range parameter and the horizontal axis representing the inference step size or the corresponding physical time.

[0111] Step 144: Generate a description of the process tolerance degradation trend based on the slope change characteristics of the predicted expansion trajectory.

[0112] Piecewise linear fitting or local weighted regression smoothing is applied to the predicted extended trajectory to eliminate high-frequency noise fluctuations and highlight its macroscopic trend characteristics. The first derivative of the smoothed trajectory in each local time interval is calculated, representing the degradation rate of the process tolerance index during that period. By scanning the degradation rate sequence across the entire time axis, the inflection point where the degradation rate changes from relatively gradual to a sharp increase is identified. The inflection point is identified using curvature extremum detection or second derivative zero-crossing point detection algorithms. When the increase in degradation rate significantly exceeds its historical fluctuation range near a certain time point, that time point is marked as the degradation acceleration inflection point. Multiplying the inference step number corresponding to the degradation acceleration inflection point by the inference step time yields the predicted operating time of the degradation acceleration inflection point.

[0113] Simultaneously, qualitative trend indicators are assigned to the degradation status of each workstation based on the absolute value and trend of the degradation rate. When the degradation rate is below the first preset rate level, it is identified as slow degradation; when the degradation rate is between the first and second preset rate levels, it is identified as accelerated degradation; and when the degradation rate is above the second preset rate level, it is identified as rapid degradation. The first and second preset rate levels are determined based on the statistical distribution quantiles of the degradation rate under historical normal operating conditions for that workstation. The description of the process tolerance degradation trend is jointly constituted by the set of predicted operating times for the degradation acceleration inflection point corresponding to each workstation and the set of qualitative trend indicators for the degradation rate.

[0114] Step 145: Based on the correlation description between the gain coefficient offset direction and the tendency of potential resonance risk in subsequent workstations in the predictive logic structure, and the propagation attenuation characteristics of the transient impact behavior component in the transmission path between multiple workstations on the production line, the superposition effect of process impact in the chain transmission process of multiple workstations is simulated in the virtual production line simulation environment.

[0115] In the virtual production line simulation environment, each virtual behavior simulation unit may generate transient impact behavior components within each simulation step. When a virtual behavior simulation unit generates an impact, the impact propagates to downstream virtual behavior simulation units topologically connected to it according to the amplitude attenuation coefficient sequence and phase delay accumulation rule defined in step 1414. The propagation process proceeds bidirectionally along the material flow direction and structural connection path of the production line, with a focus on the forward propagation from the chip mounting station to the insertion station and then to the soldering station. When the shock wave reaches the downstream virtual behavior simulation unit, it intersects with the shock wave that the unit itself may generate within the current step on the time axis.

[0116] If the time difference between two shock waves arriving at the same virtual behavior simulation unit is less than the unit's current predicted recovery time parameter, the two shock waves overlap in the time domain, resulting in a superposition effect. The superposition effect is calculated by adding the instantaneous amplitudes. At each sampling point within the overlapping period, the instantaneous amplitude of the synthesized shock signal is equal to the sum of the instantaneous amplitude of the upstream transmitted shock signal and the instantaneous amplitude of the shock signal generated by the unit itself. If multiple upstream shock sources simultaneously transmit to the same unit, the instantaneous amplitude of the synthesized shock signal is the sum of the instantaneous amplitudes contributed by all shock sources.

[0117] Step 146: Identify the workstation combination areas that meet the resonance risk activation conditions under the superposition effect, map the workstation combination areas to the actual physical layout coordinates of the production line, and generate a potential resonance workstation area distribution description.

[0118] The resonance risk triggering condition is defined as follows: within multiple consecutive simulation steps, the instantaneous amplitude of the synthesized impact signal of a virtual behavior simulation unit exceeds a preset resonance risk triggering threshold, and the phase difference of the impact response of the virtual behavior simulation unit and its adjacent upstream and downstream virtual behavior simulation units remains constant or exhibits a regular phase-locked relationship within multiple consecutive simulation steps. Virtual behavior simulation units that meet the above conditions are marked as resonance risk units.

[0119] On the production line topology diagram, each resonant risk unit is used as a seed node to search outwards for adjacent units with strong coupling relationships. The strength of the coupling relationship is determined by whether the effective coupling strength coefficient defined in step 1413 is higher than a preset coupling threshold. These units, along with the seed node, are grouped into a resonant risk workstation combination area. For each identified resonant risk workstation combination area, the system maps its node number in the virtual production line topology back to the actual workstation number sequence in the physical production line, and reads the workstation coordinate information from the physical layout drawing of the production line to determine the boundary coordinates of the minimum outer rectangle or polygon of the risk area, extracting significant spatial reference markers near the boundary. The distribution description of potential resonant workstation areas is output in list form, with each item in the list corresponding to a resonant risk area, including the workstation number sequence covered by the area and the physical location reference marker of the area boundary.

[0120] Step 150: Generate a production line control instruction set based on the process tolerance degradation trend description and the potential resonance station area distribution description. The production line control instruction set is used to adjust the driving force input characteristic data of the corresponding actuator or the action sequence configuration data between stations.

[0121] After predicting the future degradation trend of process tolerance and spatially locating potential resonance risk areas, the electronic component production line optimization system enters the decision-making and control instruction generation stage. The generation of control instructions follows the basic principles of prioritizing the handling of urgent risks and coordinating the adjustment of related workstations. The instructions cover two aspects: adjusting the input characteristic data of the actuator driving force and adjusting the timing configuration data of actions between workstations.

[0122] Step 151: Analyze the predicted operating time and qualitative trend identifier of the inflection point of process tolerance degradation acceleration corresponding to each station in the description of process tolerance degradation trend, and generate a priority ranking table for the adjustment of dynamic response parameters of each station actuator.

[0123] The description of the process tolerance degradation trend includes the predicted operating time of the degradation acceleration inflection point and a qualitative trend indicator of the degradation rate for each workstation. A weighted scoring mechanism is used to calculate the urgency score for each workstation. Different baseline scores are assigned to different qualitative trend indicators of degradation rates: the highest baseline score corresponds to the indicator of rapid degradation, the medium baseline score corresponds to the indicator of accelerated degradation, and the lowest baseline score corresponds to the indicator of slow degradation.

[0124] Furthermore, a time proximity factor is introduced to weight and correct the baseline score. The time proximity factor is calculated by dividing the preset maximum prediction duration by the predicted operation time of the degradation acceleration inflection point. The final control urgency score for each workstation is equal to its baseline score multiplied by the time proximity factor. All workstations are sorted in descending order of control urgency score to generate a priority ranking table for dynamic response parameter adjustments. The workstation with the highest score in the ranking table has the highest adjustment priority, and the system will prioritize generating control commands for that workstation.

[0125] Step 152: Based on the workstation number sequence covered by the risk area in the potential resonance workstation area distribution description, identify multiple actuators located in the same resonance risk workstation combination area, and generate a collaborative adjustment scheme for the production line cycle time configuration parameters for the multiple actuators.

[0126] When the workstation number sequence contains more than one workstation, it indicates a significant impact transmission and superposition effect among the actuators of multiple workstations within the area. The core idea of ​​the solution is to disrupt the time condition for impact energy superposition by staggering impact events that might otherwise be synchronous or close in phase. For the u-th actuator within the risk area, a phase stagger adjustment is added to its original action sequence configuration.

[0127] The phase offset adjustment is calculated using an equal-interval allocation strategy, equal to one reference offset period multiplied by u and then divided by the total number of actuators in the region. The reference offset period is a fraction of the production cycle time. Through this equal-interval offset arrangement, multiple impact events that might have overlapped in time are distributed across different moments in the production cycle time, maximizing the time interval between any two impact events. The coordinated adjustment scheme for production line cycle time configuration parameters is recorded in tabular form, with each row corresponding to the identifier of an actuator and the phase offset adjustment amount to be applied.

[0128] Step 153: For the actuator corresponding to the highest priority workstation in the power response parameter adjustment priority sorting table, generate a first control instruction. The first control instruction includes the amplitude scaling ratio parameter of the timing change waveform of the driving torque of the actuator and the compensation correction amount of the control instruction response delay parameter.

[0129] The first control command is a fine-tuning command for a single actuator, the goal of which is to reduce the impact intensity generated during operation or improve its ability to suppress impact by modifying the driving force input characteristics of the actuator.

[0130] The first control command is generated based on the description of the process tolerance degradation trend of this station and the continuous change profile curve and phase drift trajectory curve of the power transmission gain extracted in step 121. The amplitude scaling parameter is used to compress or expand the amplitude of the overall driving torque timing waveform. When the degradation rate qualitative trend is identified as rapid degradation or accelerated degradation, the amplitude scaling parameter is taken as a value smaller than the reference value to reduce the peak value and rate of change of the driving torque. The compensation correction amount of the control command response delay parameter is calculated based on the current value and change trend of the phase drift trajectory curve of the cumulative hysteresis phase offset extracted in step 121 over time. The compensation correction amount is equal to the current cumulative phase hysteresis value divided by the angular frequency corresponding to the main motion frequency. The compensation correction amount is given in the form of a time value and written into the feedforward control parameters of the servo drive.

[0131] Step 154: For the actuators involved in the production line cycle time configuration parameter collaborative adjustment scheme, generate a second control command. The second control command includes the cycle time fine-tuning parameter of the speed fluctuation profile curve in the displacement response waveform of the corresponding actuator and the timing offset of the action trigger signal between adjacent workstations.

[0132] The second control command is a coordinated adjustment command for multiple actuators, aiming to adjust the timing relationship between these actuators. For each actuator involved in the production line cycle time configuration parameter coordinated adjustment scheme, the required timing offset is directly read from the scheme. The cycle time fine-tuning parameter is calculated accordingly based on the magnitude and direction of the timing offset. If the timing offset of one actuator is positive (i.e., the action is delayed), its cycle time fine-tuning parameter is set to a value slightly less than the baseline value to slightly compress its action cycle and absorb the time occupied by the delay. If the timing offset is negative (i.e., the action is advanced), the cycle time fine-tuning parameter is set to a value slightly greater than the baseline value. The adjustment range of the cycle time fine-tuning parameter is constrained by the upper limit of the equipment's physical capabilities; the system automatically checks whether the adjustment range exceeds the preset allowable range when generating the command.

[0133] Step 155: Integrate the first control instruction and the second control instruction according to the workstation arrangement order and instruction effective timing requirements to generate a production line control instruction set. The production line control instruction set is used to update the operating parameter configuration of each actuator in parallel.

[0134] First, the control commands for all involved actuators are sorted according to the arrangement of workstations in the material flow direction of the production line, with commands from upstream workstations preceding those from downstream workstations. For each actuator, if both a first control command and a second control command exist, these two commands are merged into a single composite command message, which contains a complete list of updated parameters for that actuator. An effective timestamp is added to each command message. The calculation of the effective timestamp must take into account the total delay time required from the system issuing the command to the controller receiving and parsing it, and then to the actuator completing the current action cycle.

[0135] The system selects a future full cycle boundary as the unified effective time for all instructions to ensure that parameter switching occurs during action intervals. Finally, all instruction messages with effective timestamps are packaged according to workstation order to generate a production line control instruction set, which is then distributed in parallel by the production line control system to the servo drivers or programmable logic controllers of each actuator to update the corresponding actuator's operating parameter configuration.

[0136] After step 150, the method further includes steps 160-190.

[0137] Step 160: Collect dynamic response behavior update records and workstation process tolerance status update records within a preset adjustment response period after the production line control command set is executed, and map the dynamic response behavior update records and workstation process tolerance status update records to the current status parameter description field of each virtual behavior simulation unit in the digital twin space.

[0138] To evaluate the effectiveness of the control instructions and establish a closed-loop optimization, this step collects and updates data after the control instructions are executed. The preset adjustment response period is set to a continuous operating time after the control instructions take effect. The data collection and mapping methods are the same as those described in steps 110 and 120. The updated data is used to refresh the state parameter description fields of the virtual behavior simulation units in the digital twin space, ensuring that the state of the digital twin space is synchronized with the state of the physical production line after the control measures are implemented.

[0139] Step 170: In the digital twin space, using the monotonic mapping relationship between the gain coefficient offset rate and the tolerance state deviation expansion trend in the predictive logic structure as a reference, the deviation comparison and separation processing based on the steady-state behavior performance component is performed on the update sequence of the driving force input and motion output transmission relationship characterized by the dynamic response behavior update record, and the control response offset feature sequence used to describe the degree of deviation between the mechanism behavior recovery trajectory and the expected recovery trajectory after the intervention of the control command is obtained.

[0140] This step aims to quantitatively evaluate the control effect. First, the steady-state behavior component separated in step 120 is used as the baseline behavior template. Then, a new update sequence of the driving force input and motion output transmission relationship is extracted from the dynamic response behavior update record. The updated sequence and the baseline behavior template are compared point by point under the same driving force input condition. The residual sequence of the two in motion output (such as acceleration response) is calculated. This residual sequence is the control response offset feature sequence, which describes the change over time in the deviation of the actual behavior recovery trajectory of the mechanism from the ideal expected recovery trajectory after the intervention of the control command.

[0141] Step 180: Perform a time-series change synergy determination operation on the fluctuation direction change characteristics of the control response offset feature sequence and the offset direction change characteristics of the peak transient amplitude update sequence of the station assembly contact force waveform curve in the station process tolerance state update record, and locate the period of divergence between the fluctuation direction change characteristics and the offset direction change characteristics.

[0142] The coordination determination operation employs a sign consistency test. First-order difference sign sequences of the control response offset feature sequence and peak transient amplitude update sequence are extracted separately. Then, the Hamming distance or dot product of the two sign sequences is calculated within a sliding time window. When the correlation between the sign sequences changes from positive to negative or remains negative within multiple consecutive sliding time windows, a consistency deviation is determined. The start and end times of this state are recorded as the deviation duration.

[0143] Step 190: Generate proactive compensation adjustment parameters based on the duration characteristics and amplitude variation characteristics of the divergence duration period to correct the amplitude variation envelope curvature and phase advance reference of the subsequent command waveform of the driving force input characteristic data in the production line control command set.

[0144] The longer the divergence duration and the larger the divergence amplitude, the more significant the deviation of the current control strategy. The proactive compensation adjustment parameters are calculated as follows: the duration of the divergence duration is multiplied by an integral gain coefficient and added to the amplitude gradient envelope curvature parameter of the current command waveform to adjust the falling or rising edge shape of the subsequent command waveform; simultaneously, the average slope of the divergence amplitude gradient characteristic is multiplied by a proportional gain coefficient to correct the phase lead reference of subsequent commands, thereby advancing or delaying the effective time of the commands. The corrected parameters are used to update the production line control command set generated in the next round.

[0145] After step 150, the method further includes steps 210-250.

[0146] Step 210: Load multiple sets of simulated initial operating condition parameter combinations consisting of different initial production cycle configurations and different material batch attributes into the digital twin space, and drive the digital twin space to generate corresponding dynamic response behavior simulation samples and tolerance state simulation samples for each set of simulated initial operating condition parameter combinations.

[0147] To enhance the robustness and universality of the predictive logic structure, the method also includes an offline verification and correction process. Multiple sets of initial simulation parameters are pre-configured in the digital twin space. Each set includes a set production cycle configuration parameter (e.g., main production line speed) and material batch attribute parameters (e.g., the range of plating friction coefficients for component pins). For each parameter set, the simulation engine in the digital twin space is driven to generate simulated samples of dynamic response behavior and tolerance states under that condition.

[0148] Step 220: Perform the unsupervised behavior mode separation process on each group of simulated dynamic response behavior samples to separate the corresponding simulated steady-state behavior performance components and simulated transient impact behavior performance components.

[0149] It is understandable that step 220 is executed in the same way as step 120 and its sub-steps.

[0150] Step 230: Input the simulated transient impact behavior components of each group and the simulated tolerance state at the corresponding time into the predictive logic structure for deduction processing, and generate a set of sample descriptions of the decline trend of simulated process tolerance and a set of sample descriptions of the distribution of simulated potential resonance work station areas corresponding to each group of simulated initial working condition parameter combinations.

[0151] It is understandable that step 230 is executed in the same way as step 140 and its sub-steps.

[0152] Step 240: Perform cross-condition scenario comparison analysis on the distribution concentration difference characteristics of the deterioration acceleration inflection point prediction operation time in the sample set describing the decline trend of the simulated process tolerance and the overlapping offset characteristics of the risk area covering the work station number sequence in the sample set describing the distribution of the simulated potential resonance work station area, to obtain the characterization stability evaluation characteristics of the correlation description between the gain coefficient offset direction in the predictive logic structure and the tendency of subsequent work station potential resonance risk activation.

[0153] The distribution concentration difference is measured by calculating the variance or interquartile range of the predicted operation time for the deterioration acceleration inflection point under each group of operating conditions. A smaller variance indicates more stable prediction results from the predictive logic structure under different operating conditions. The overlap offset is measured by calculating the Jaccard similarity coefficient between the risk area workstation number sequences generated under different operating conditions. A higher Jaccard similarity coefficient indicates that the identification results of the resonance area are less affected by changes in operating conditions. Combining these two indicators, a stability evaluation feature value, Stabindex, is calculated. Stabindex is calculated as follows: Stabindex equals the mean of the Jaccard similarity coefficients divided by the normalized value of the variance of the predicted operation time.

[0154] Step 250: Based on the stability evaluation characteristics, the attenuation coefficient sequence described by the propagation attenuation characteristics in the predictive logic structure is modified to be oriented towards the range of operating condition variations. The modified predictive logic structure is determined as the common behavior transmission description benchmark for subsequent multi-operating condition deduction in the digital twin space.

[0155] If the stability evaluation feature value Stabindex is lower than the preset stability threshold, it indicates that the original attenuation coefficient sequence is sensitive to changes in operating conditions. In this case, a correction process is initiated. The correction method is as follows: each attenuation coefficient attcoefk in the attenuation coefficient sequence is multiplied by an adjustment factor, which is 1 plus the difference between Stabindex and the stability threshold multiplied by a negative correction coefficient. The corrected attenuation coefficient sequence makes the predicted resonance risk region distribution more consistent when facing different operating conditions, improving the model's generalization ability. The corrected predictive logic structure is saved as a globally shared behavioral transfer description benchmark in the digital twin space for use in all subsequent online simulation tasks.

[0156] This application's embodiments map synchronously acquired actuator dynamic response behavior records and workstation process tolerance status records to a digital twin space. Unsupervised behavior mode separation processing is applied to the actuator dynamic response behavior records to extract time-synchronized steady-state behavior components and transient impact behavior components. Then, within the digital twin space, semantic association analysis of the transient impact behavior components and workstation process tolerance status records is performed to construct a predictive logical structure. This predictive logical structure enables virtual actors to deduce the evolution trend of process tolerance status records and the superposition effect of transient impact behavior components between workstations. This generates descriptions of process tolerance degradation trends and potential resonance workstation area distributions to guide the generation of production line control instruction sets. This method achieves cross-level association from macroscopic production line behavior analysis to microscopic dynamic response mapping, effectively improving the accuracy of process state prediction and control in continuous production cycles for electronic component production lines.

[0157] Based on the above, those skilled in the art can combine the transparent clock and delay measurement mechanism defined in the Industrial Ethernet Precision Time Protocol to establish a high-precision synchronization link between distributed acquisition nodes, constrain the trigger time deviation of each sensor to the sub-microsecond range, thereby ensuring that the two heterogeneous data, namely the dynamic response of the actuator and the process tolerance status, have a point-to-point alignment reference in the timestamp dimension.

[0158] In the unsupervised behavior mode separation stage, the variational mode decomposition algorithm framework can be combined. By constructing a variational optimization model with the goal of minimizing the sum of the estimated bandwidths of each mode, and using the alternating direction multiplier method to iteratively update the center frequency and analytic signal of each intrinsic mode function in the frequency domain, the original acceleration response time series can be adaptively decoupled into multiple oscillating components with different center frequencies and finite bandwidths. The built-in Wiener filtering noise reduction characteristics of this algorithm can effectively avoid the mode aliasing phenomenon, so that the separated low-frequency components accurately represent the long-term slow variation law of the stiffness and damping parameters of the transmission system, while the high-frequency components completely retain the impact attenuation envelope shape at the moment of mechanism start-up and shutdown.

[0159] For quantitative assessment of the strength of mutual information correlation, an entropy estimation method based on K-nearest neighbor statistics can be introduced. This method searches for the Kth nearest neighbor in the neighborhood of the sample point in the joint state space and each edge state space and calculates its geometric distance. Combined with the recursive relationship of the Digamma function, it derives the unbiased estimate of the mutual information between variables, thereby robustly measuring the degree of statistical dependence between each behavioral mode component and the contact force waveform curve without relying on the prior assumption of probability density.

[0160] In the predictive logic structure construction stage, the general approximation capability of the adaptive neurofuzzy inference system can be relied upon to automatically identify cluster centers in the input feature space composed of the initial amplitude and recovery time using the subtractive clustering algorithm to generate fuzzy rule antecedents. Then, through the global optimal solution of the linear coefficients of the rule consequent by the least squares estimator in the hybrid learning algorithm and the local fine adjustment of the shape parameter of the membership function by the gradient descent method, the nonlinear mapping function describing the influence of impact behavior on the transmission of process tolerance state can be iteratively identified.

[0161] To quantify the attenuation characteristics of impact transmission between workstations, the production line connection links can be equivalently represented as flexible interfaces with complex stiffness based on the comprehensive theory of mechanical impedance and admittance. The energy transfer efficiency of the interface can be inverted by calculating the Pearson correlation coefficient of the envelope sequence of transient impact behavior components of upstream and downstream workstations. The higher the correlation coefficient, the smaller the corresponding amplitude attenuation coefficient value, thereby constructing a chain attenuation coefficient sequence that conforms to the actual topological constraints of the production line.

[0162] In the control command generation stage, the phase advance correction principle in the feedforward compensation architecture of the servo control system can be used. Based on the time constant obtained by dividing the cumulative phase lag obtained by the identification by the corresponding angular frequency of the main motion frequency, an equal amount of time advance offset is pre-inserted in the position command planner to counteract the response lag effect caused by the combined flexible deformation of the transmission chain and electrical delay.

[0163] Please see Figure 2 The figure is a schematic diagram of the basic structure of an electronic component production line optimization system 200 provided in an embodiment of this application. The electronic component production line optimization system 200 includes: a processor 201; a storage device 202 storing a computer program 2020; and a network interface 203 for providing network communication functions. When the computer program 2020 is executed by the processor 201, the processor 201 implements any of the aforementioned electronic component production line optimization methods using digital twins.

[0164] Please see Figure 3 This application provides a functional block diagram of an electronic component production line optimization device, which includes: The production line record acquisition module is used to acquire the actuator dynamic response behavior record and the workstation process tolerance status record synchronously collected within a preset continuous production cycle of the target electronic component production line. The actuator dynamic response behavior record includes a description of the instantaneous transmission characteristics between the driving force input and motion output of each actuator. The behavior pattern separation module is used to map the dynamic response behavior record of the actuator and the process tolerance state record of the workstation to a preset digital twin space, and perform unsupervised behavior pattern separation processing on the dynamic response behavior record of the actuator in the digital twin space to separate the steady-state behavior performance component and the transient impact behavior performance component that are synchronized with the process tolerance state record of the workstation in time. The semantic association parsing module is used to perform behavioral semantic association parsing on the transient impact behavior performance component and the process tolerance state record of the workstation at the same time in the digital twin space, and to construct a predictive logical structure describing the impact of transient impact behavior on the transmission of process tolerance state. The digital twin simulation module is used to apply the predictive logic structure to the virtual behavior of the actuator in the digital twin space, to simulate the evolution trend of the process tolerance status record of the workstation in subsequent continuous production cycles and the superposition effect of the transient impact behavior component transmitted between workstations, and to generate a description of the process tolerance decay trend and a description of the potential resonance workstation area distribution. The control instruction generation module is used to generate a set of production line control instructions based on the description of the process tolerance degradation trend and the description of the potential resonance station area distribution. The set of production line control instructions is used to adjust the driving force input characteristic data of the corresponding actuator or the action sequence configuration data between stations.

[0165] Based on the above, a readable storage medium is provided, on which a program or instructions are stored, and when the program or instructions are executed by a processor, the steps of the above method are implemented.

[0166] Furthermore, it should be noted that this application also provides a computer program product, which may include a computer program that can be stored in a computer-readable storage medium. The processor of the electronic component production line optimization system reads the computer program from the computer-readable storage medium, and the processor can execute the computer program, causing the electronic component production line optimization system to perform the aforementioned... Figure 1 The methods described in the corresponding embodiments are already known, and therefore will not be repeated here. Furthermore, the beneficial effects of using the same method will also not be repeated. For technical details not disclosed in the computer program product embodiments related to this application, please refer to the description of the method embodiments of this application.

[0167] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems or apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple, and relevant parts can be referred to the method section.

Claims

1. A method for optimizing electronic component production lines using digital twins, characterized in that, The method includes: Acquire the actuator dynamic response behavior record and the workstation process tolerance status record synchronously collected within a preset continuous production cycle of the target electronic component production line. The actuator dynamic response behavior record includes a description of the instantaneous transmission characteristics between the driving force input and motion output of each actuator. The dynamic response behavior record of the actuator and the process tolerance status record of the workstation are mapped to a preset digital twin space. Unsupervised behavior mode separation processing is performed on the dynamic response behavior record of the actuator in the digital twin space to separate the steady-state behavior performance component and the transient impact behavior performance component that are time-synchronized with the process tolerance status record of the workstation. Within the digital twin space, the behavioral semantic association analysis of the transient impact behavior component and the process tolerance status record of the workstation at the same time is performed to construct a predictive logical structure describing the impact of transient impact behavior on the transmission of process tolerance status. The predictive logic structure is applied to the virtual behavior of the actuator in the digital twin space to deduce the evolution trend of the process tolerance status record of the workstation in subsequent continuous production cycles and the superposition effect of the transient impact behavior component transmitted between workstations, thereby generating a description of the process tolerance decay trend and a description of the distribution of potential resonance workstation areas. Based on the description of the process tolerance degradation trend and the description of the potential resonance workstation area distribution, a set of production line control instructions is generated. The set of production line control instructions is used to adjust the driving force input characteristic data of the corresponding actuator or the action sequence configuration data between workstations.

2. The method according to claim 1, characterized in that, The process involves mapping the actuator's dynamic response behavior record and the workstation's process tolerance state record to a preset digital twin space. Within this digital twin space, unsupervised behavior mode separation processing is performed on the actuator's dynamic response behavior record to separate the steady-state behavior component and transient impact behavior component that are time-synchronized with the workstation's process tolerance state record. This includes: The instantaneous transmission characteristics of each actuator in the dynamic response behavior record of the actuator are analyzed, and the continuous change profile curve of the change in the dynamic transmission gain over time and the phase drift trajectory curve of the cumulative value of the hysteresis phase offset over time are extracted. In a pre-defined digital twin space, virtual behavior simulation units corresponding to each actuator in the physical production line are established. The continuously changing contour curve and the phase drift trajectory curve are respectively loaded into the motion transmission attribute description model of the corresponding virtual behavior simulation unit to generate a virtual mechanism dynamic behavior representation that is synchronized with the behavior of the physical actuator in real time. An unsupervised blind separation operation based on self-similarity metric is performed on the behavior representation signal sequence output by the virtual mechanism dynamic behavior representation body to decompose the behavior representation signal sequence into several behavior pattern components with different fluctuation characteristics. Calculate the mutual information correlation strength between each of the behavioral pattern components and the station assembly contact force waveform curve in the station process tolerance state record; Behavioral pattern components with mutual information correlation strength lower than the first preset judgment level are classified as steady-state pattern component candidate set, and behavioral pattern components with mutual information correlation strength higher than the second preset judgment level are classified as transient impact pattern component candidate set. By filtering the candidate sets of steady-state mode components and the candidate sets of transient impact mode components, the steady-state behavior performance components and the transient impact behavior performance components are obtained and time-synchronized.

3. The method according to claim 2, characterized in that, The step of filtering the candidate sets of steady-state mode components and the candidate sets of transient impact mode components to obtain the steady-state behavior performance components and the transient impact behavior performance components, and then performing time synchronization processing, includes: The behavioral mode components in the steady-state mode component candidate set are clustered and merged according to their fluctuation frequency range to obtain the steady-state behavioral performance components that characterize the long-term stable operation state of the actuator. The steady-state behavioral performance components include the baseline drift trend parameter of the power transmission gain and the low-frequency cumulative change characteristics of the phase drift. Each behavioral mode component in the transient impact mode component candidate set is screened based on the time overlap of its peak transient amplitude parameter with the waveform curve of the assembly contact force at the workstation, to obtain the transient impact behavior performance component that characterizes the instantaneous change in the mechanism behavior caused by the process impact event. The transient impact behavior performance component includes the impact response amplitude attenuation envelope and the phase reset dynamic process characteristics after the impact disturbance. The steady-state behavior component and the transient impact behavior component are respectively calibrated with the corresponding workstation process tolerance status record on the time axis, so that the steady-state behavior component and the transient impact behavior component are synchronized with the workstation process tolerance status record in time.

4. The method according to claim 1, characterized in that, The process involves performing behavioral semantic association analysis on the transient impact behavior components and the process tolerance status record of the workstation at the same time within the digital twin space, constructing a predictive logical structure describing the impact of transient impact behavior on the transmission of process tolerance status, including: Extract the initial amplitude parameter of the impact response amplitude attenuation envelope and the recovery time parameter of the phase reset dynamic process characteristics after impact disturbance from the transient impact behavior performance components. Extract the peak transient amplitude parameter and the time-domain fluctuation range parameter of the assembly contact force waveform curve synchronized with the time of each process impact event from the process tolerance status record of the workstation, and generate the process tolerance instantaneous deviation record pair. A process shock risk correlation model is established within the digital twin space. The process shock risk correlation model uses the initial amplitude parameter and the recovery time length parameter as the set of input variables, and the peak transient amplitude parameter and the time domain fluctuation range parameter as the set of output observation variables. The co-evolutionary transitive relationship between the set of input variables and the set of output observation variables is learned by unsupervised association rule mining, and the transfer function description of the input variables to the output observation variables is obtained. Based on the process tolerance instantaneous deviation record pairs corresponding to multiple sets of process shock events occurring sequentially during a continuous operation period, the shift direction and shift rate of the gain coefficient described by the transfer function as a function of continuous operation time are calculated. Based on the offset direction and offset rate of the gain coefficient, a predictive logic structure is constructed to describe the impact of transient impact behavior on the transmission of process tolerance status. The predictive logic structure includes a monotonic mapping relationship between the gain coefficient offset rate and the trend of the expansion of tolerance status deviation, as well as a correlation description between the gain coefficient offset direction and the tendency of potential resonance risk in subsequent workstations.

5. The method according to claim 4, characterized in that, The predictive logic structure also includes a description of the propagation attenuation characteristics of the transient impact behavior component as it is transmitted between multiple workstations on the production line. The propagation attenuation characteristics are determined based on the correlation analysis results between the transient impact behavior components corresponding to adjacent workstations.

6. The method according to claim 1, 2, or 4, characterized in that, The process involves applying the predictive logic structure to the virtual actors of the actuators within the digital twin space to deduce the evolution trend of the process tolerance status records of the workstations in subsequent continuous production cycles and the superposition effect of the transient impact behavior components transmitted between workstations, generating a description of the process tolerance degradation trend and a description of the potential resonance workstation area distribution, including: The predictive logic structure is loaded into the virtual production line simulation environment in the digital twin space. The virtual production line simulation environment includes a set of virtual behavior simulation units corresponding to each workstation and each actuator in the physical production line, as well as a power behavior transmission path model between the virtual behavior simulation units. Based on the monotonic mapping relationship between the gain coefficient offset rate and the tolerance state deviation expansion trend in the predictive logic structure, the transient impact behavior performance components of each actuator in the virtual production line simulation environment are predicted during a preset continuous operation period, resulting in the predicted initial amplitude sequence and predicted recovery time sequence of the transient impact behavior performance components during the preset continuous operation period. The predicted initial amplitude sequence and the predicted recovery duration sequence are input into the transfer function description in the predictive logic structure to deduce the predicted value sequence of the peak transient amplitude parameter of the assembly contact force waveform curve of each workstation within the preset continuous operation period and the predicted expansion trajectory of the time domain fluctuation range parameter of the assembly positioning feedback deviation value. Based on the slope change characteristics of the predicted expansion trajectory, a process tolerance degradation trend description is generated. The process tolerance degradation trend description includes the predicted operating time corresponding to the deterioration acceleration inflection point of the process tolerance index of each station and the qualitative trend indicator of the deterioration rate. Based on the correlation description between the gain coefficient offset direction and the tendency of potential resonance risk in subsequent workstations in the predictive logic structure, and the propagation attenuation characteristics of the transient impact behavior component in the transmission path between multiple workstations on the production line, the superposition effect of process impact in the chain transmission process of multiple workstations is simulated in the virtual production line simulation environment. Identify workstation combination areas that meet the resonance risk triggering conditions under the superposition effect, map the workstation combination areas to the actual physical layout coordinates of the production line, and generate a potential resonance workstation area distribution description. The potential resonance workstation area distribution description includes the workstation number sequence covered by the risk area and the physical location reference identifier of the risk area boundary.

7. The method according to claim 6, characterized in that, The step of loading the predictive logic structure into the virtual production line simulation environment in the digital twin space includes: A virtual production line simulation environment is instantiated in the digital twin space, and a snapshot of the state parameters of each virtual behavior simulation unit at the current moment is inherited from the digital twin space. The snapshot of the state parameters includes the steady-state behavior performance component corresponding to each actuator and the transient impact behavior performance component corresponding to the most recent process impact event. The monotonic mapping relationship between the gain coefficient offset rate and the expansion trend of the tolerance state deviation in the predictive logic structure is described and converted into the state transition rule of the process tolerance index of each station in the virtual production line simulation environment as the simulation step size is updated. The correlation description between the gain coefficient offset direction in the predictive logic structure and the tendency of potential resonance risk in subsequent workstations is converted into a coupling strength adjustment rule for the dynamic behavior transmission between adjacent workstations in the virtual production line simulation environment. The propagation attenuation characteristics of the transient impact behavior component in the predictive logic structure, which is transmitted between multiple workstations on the production line, are transformed into the amplitude attenuation coefficient sequence and phase delay accumulation rule when the impact response is transmitted between multi-level virtual behavior simulation units in the virtual production line simulation environment. In the virtual production line simulation environment, the simulation clock is started, and the state transition rule, the coupling strength adjustment rule, and the amplitude attenuation coefficient sequence and phase delay accumulation rule are executed cyclically according to the preset simulation step size, driving the behavior state of each virtual behavior simulation unit to evolve forward in sequence.

8. The method according to claim 6, characterized in that, The method describes the evolution prediction of the transient impact behavior components of each actuator in the virtual production line simulation environment within a preset continuous operation period, based on the monotonic mapping relationship between the gain coefficient offset rate and the tolerance state deviation expansion trend in the predictive logic structure. This yields the predicted initial amplitude sequence and predicted recovery time sequence of the transient impact behavior components within the preset continuous operation period, including: Within each simulation step of the virtual production line simulation environment, the initial amplitude parameters and recovery time length parameters of the transient impact behavior performance components corresponding to each actuator in the previous simulation step are obtained; Based on the monotonic mapping relationship between the gain coefficient offset rate and the expansion trend of the tolerance state deviation, the incremental adjustment direction and adjustment magnitude of the initial amplitude parameter of the transient impact behavior performance component within the current simulation step are determined relative to the previous simulation step. The incremental adjustment direction is in the same direction as the offset direction of the gain coefficient. The predicted initial amplitude parameters of the transient impact behavior manifestation components within the current simulation step are updated according to the incremental adjustment direction and adjustment magnitude. The updated predicted initial amplitude parameters are then applied to the impact response amplitude attenuation envelope generation process of the virtual behavior simulation unit within the current simulation step. Based on the relative relationship between the updated predicted initial amplitude parameter and the baseline drift trend parameter of the power transmission gain in the steady-state behavior component, the predicted recovery time length parameter of the transient impact behavior component within the current simulation step is adjusted so that the predicted recovery time length parameter monotonically increases as the predicted initial amplitude parameter increases. The predicted initial amplitude parameter and the predicted recovery time length parameter generated at each simulation step are stored in the order of the simulation step to generate the predicted initial amplitude sequence and the predicted recovery time length sequence of the transient impact behavior performance components within a preset continuous operation period.

9. The method according to any one of claims 1-4, characterized in that, The process of generating a set of production line control instructions based on the description of the process tolerance degradation trend and the description of the potential resonance workstation area distribution includes: The qualitative trend identifiers of the predicted operation time and degradation rate of the process tolerance degradation acceleration inflection point corresponding to each station in the process tolerance degradation trend description are analyzed to generate a priority ranking table for the adjustment of dynamic response parameters of each station actuator. The station with the degradation rate qualitative trend identifier as a rapid degradation identifier and the earliest predicted operation time is arranged in the highest adjustment priority. Based on the workstation number sequence covered by the risk area in the description of potential resonance workstation area distribution, multiple actuators located in the same resonance risk workstation combination area are identified, and a production line cycle time configuration parameter collaborative adjustment scheme is generated for the multiple actuators. The production line cycle time configuration parameter collaborative adjustment scheme includes the phase offset adjustment amount of the action timing of each actuator. For the actuator corresponding to the highest priority workstation in the power response parameter adjustment priority sorting table, a first control instruction is generated. The first control instruction includes the amplitude scaling ratio parameter of the timing change waveform of the driving torque of the actuator and the compensation correction amount of the control instruction response delay parameter. For the actuators involved in the production line cycle time configuration parameter collaborative adjustment scheme, a second control instruction is generated. The second control instruction includes the cycle time frequency fine-tuning parameter of the speed fluctuation profile curve in the displacement response waveform of the corresponding actuator and the timing offset of the action trigger signal between adjacent workstations. The first control instruction and the second control instruction are integrated according to the workstation arrangement order and the instruction effective timing requirements to generate a production line control instruction set. The production line control instruction set is used to update the operating parameter configuration of each actuator in parallel.

10. An electronic component production line optimization system, characterized in that, include: A processor; a storage device having a computer program stored thereon; a network interface for providing network communication functions; when the computer program is executed by the processor, the processor enables the processor to implement the electronic component production line optimization method using digital twins as described in any one of claims 1-9.

Images