A digital-twin-based SMT whole-line three-dimensional dynamic modeling and simulation method and system

Abstract

The present application relates to the technical field of digital twinning and intelligent manufacturing simulation, in particular to a SMT whole line three-dimensional dynamic modeling simulation method and system based on digital twinning, comprising: establishing a SMT whole line equipment three-dimensional model and a mapping relationship; uniformly marking, time synchronizing and process constraint integrating initial physical collection data, and outputting whole line state data aligned by time; singular value decomposing continuous multiple frames of state data; determining the number of reserved components in combination with rendering frame rate, generating whole line rendering data; updating the decomposition result or executing rendering mode switching through real image and virtual image matching degree checking, and outputting three-dimensional dynamic modeling simulation result; the present application reduces the causal misplacement caused by sampling frequency difference, network jitter and independent clock drift.

Description

Technical Field

[0001] This invention relates to the field of digital twin and surface mount production line simulation technology, specifically to a three-dimensional dynamic modeling and simulation method and system for an entire SMT production line based on digital twins. Background Technology

[0002] As electronic manufacturing production lines develop towards higher speeds, greater flexibility, and greater collaboration, the number of solder paste printing equipment, mounting equipment, soldering equipment, and testing equipment in surface mount assembly lines continues to increase, and the data interaction and action linkage relationships between these devices are becoming increasingly complex. To ensure the accuracy of production line debugging, operation monitoring, and anomaly analysis, it is usually necessary to establish a corresponding three-dimensional dynamic modeling simulation model for the surface mount assembly line so as to synchronously present and recreate the real production line status in a virtual scene. However, when performing 3D dynamic modeling and simulation of the entire surface mount assembly line, it is necessary to process the multi-source state data generated by different devices in a unified manner and drive the 3D model to respond continuously. Due to the differences in sampling frequency, data format and action rhythm of each device, the existing methods usually use static model superimposed state display or simple event-triggered animation for simulation. This processing method is difficult to balance the temporal consistency and action continuity of the entire line state, and is prone to problems such as virtual actions preceding physical actions, missing key motion details, or excessive rendering load. This instability not only affects the realism of the 3D simulation results, but may also reduce their effectiveness in production line debugging, status monitoring, and anomaly detection. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for three-dimensional dynamic modeling and simulation of an entire SMT production line based on digital twins, avoiding the occurrence of virtual actions preceding physical actions or the loss of key motion details, and synchronizing the actual production line operation status under limited computing resources. Specifically, the technical solution of this invention is as follows: A digital twin-based 3D dynamic modeling and simulation method for an entire SMT production line includes: Establish a 3D model of the SMT production line equipment and establish a mapping relationship between the virtual equipment status and the equipment joint parameters or key process parameters; Acquire the initial physical acquisition data of the SMT line equipment, perform time synchronization and integration based on process timing constraints, and output time-aligned line status data; Extract the overall line status data to construct multi-frame status data, perform singular value decomposition, and determine the state decomposition matrix and the proportion of each order component. Obtain the current rendering frame rate feedback result, and dynamically adjust based on the proportion of each order component and the current rendering frame rate feedback result to determine the number of retained components; The state decomposition matrix is ​​divided into primary and secondary components based on the number of retained components. The primary driving data is extracted from the primary components, and the secondary components are attenuated to generate secondary compensation data. The whole line rendering data is then fused and output. Calculate the image matching degree between the real image of the physical production line and the virtual image generated based on the whole line rendering data. Based on the relationship between the image matching degree and the preset consistency threshold, update the state decomposition matrix or execute the rendering mode switching control.

[0004] Optionally, time synchronization and integration are performed based on process timing constraints, and the output time-aligned whole-line status data includes: Add a uniform absolute time stamp to the initial physical acquisition data, and resample according to the system simulation step size to generate single-frame synchronization state data; Obtain the exit trigger time of the upstream device and the inlet trigger time of the downstream device, as well as the device transmission parameters, and calculate the theoretical arrival time. Based on the comparison between the entry trigger time and the theoretical arrival time of the downstream equipment, the process timing constraint status is determined, and interpolation compensation is performed on the abnormal state quantities with timing conflicts.

[0005] Optionally, singular value decomposition is performed to determine the state decomposition matrix and the proportion of each order of components, including: Perform mean-removal processing on multi-frame state data; Singular value decomposition is performed on the multi-frame state data after mean removal processing to extract the state decomposition matrix, singular value set, and time coefficient sequence that reflect the overall characteristics of the line, and the proportion of each order component is calculated based on the singular value set.

[0006] Optionally, the number of retained components is dynamically adjusted based on the proportion of each order component and the current rendering frame rate feedback, including: The number of candidate truncated components is determined based on the proportion of each order of components and the preset weight threshold. If the current rendering frame rate feedback result is less than the system target rendering frame rate, then the preset weight threshold and the number of candidate truncated components are reduced; if the current rendering frame rate feedback result is greater than the sum of the system target rendering frame rate and the allowed upward difference, then the preset weight threshold and the number of candidate truncated components are increased; if the current rendering frame rate feedback result is greater than or equal to the system target rendering frame rate and less than or equal to the sum of the system target rendering frame rate and the allowed upward difference, then the number of candidate truncated components remains unchanged.

[0007] Optionally, extracting the main driving data from the main components and performing attenuation processing on the secondary components to generate secondary compensation data includes: For the main components, the main driving data mapped to the equipment joint parameters or key process parameters are generated by reverse calculation. For the minor component, the characteristic coefficient value corresponding to the minor component is obtained from the time coefficient sequence generated by the singular value decomposition of multi-frame state data. Attenuation calculation is performed based on the time interval between the current time and the most recent update time of the characteristic coefficient value and the preset damping attenuation coefficient to generate minor compensation data.

[0008] Optionally, updating the state decomposition matrix or performing rendering mode switching control based on the relationship between image matching degree and preset consistency threshold includes: If the image matching degree is less than the preset consistency threshold, the historical decomposition cumulative results are discarded, and multi-frame state data is reconstructed to update the state decomposition matrix. If the image matching degree is greater than or equal to the preset consistency threshold, the current state decomposition matrix is ​​maintained, and the rendering mode switching control is executed.

[0009] Optionally, the execution of rendering mode switching control includes: Extract the singular value set generated by singular value decomposition from multiple frames of state data, and calculate the ratio of singular value extrema in the current singular value set; If the singular value-to-extreme ratio is greater than the preset ratio threshold, the system will switch to simplified rendering and motion response mode; if the singular value-to-extreme ratio is less than or equal to the preset ratio threshold, the primary and secondary processing flow will be restored.

[0010] A digital twin-based 3D dynamic modeling and simulation system for an entire SMT production line includes: The model building module is used to build a 3D model of the SMT production line equipment and establish a mapping relationship between the virtual equipment status and the equipment joint parameters or key process parameters. The data acquisition and synchronization module is used to acquire initial physical data, synchronize and integrate it in time, and output time-aligned whole-line status data. The decomposition and analysis module is used to construct multi-frame state data and perform singular value decomposition. The rendering precision control module is used to dynamically determine the number of components to be retained based on the component ratio and rendering frame rate feedback results. The data merging module is used to process the primary and secondary components according to the number of retained components, and merge and output the whole line rendering data. The virtual-real image comparison module is used to calculate the image matching degree between real and virtual images, update the state decomposition matrix, or perform rendering mode switching control.

[0011] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention establishes a three-dimensional model of the entire production line and constructs a mapping relationship between the equipment joint parameters or key process parameters and the virtual equipment status. Combined with unified absolute time stamping, resampling according to the system simulation step size, and process timing constraint verification based on upstream exit triggering, downstream inlet triggering, equipment spacing and conveyor belt speed, it outputs time-aligned whole-line status data for multi-source heterogeneous acquisition data, reducing causal misalignment caused by sampling frequency differences, network jitter and independent clock drift. 2. This invention constructs state data from multiple consecutive frames, performs mean removal processing and singular value decomposition, and extracts the state decomposition matrix, singular value set, time coefficient sequence and the proportion of each order component. It can extract the main linkage change features from the entire high-dimensional state and provide a quantitative basis for the selection of retained components in subsequent rendering. 3. This invention dynamically adjusts the preset weight threshold and the number of retained components based on the proportion of each component and the feedback result of the current rendering frame rate. It performs inverse calculation on the main components and damping attenuation compensation on the secondary components before performing same-dimensional fusion to output the whole-line rendering data. This allows for the balance of key action fidelity, weak component continuity and rendering smoothness under limited computing resources, reducing excessive rendering load and screen jump problems. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of this application and the prior art, the accompanying drawings used in the description of the embodiments and the prior art will be briefly introduced below: Figure 1 This is a flowchart illustrating a three-dimensional dynamic modeling and simulation method for an entire SMT production line based on digital twins, provided in an embodiment of this application. Figure 2 This is a schematic diagram of a module of a digital twin-based three-dimensional dynamic modeling and simulation system for an entire SMT production line, provided in an embodiment of this application. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0014] Please see Figure 1 A three-dimensional dynamic modeling and simulation method for an entire SMT production line based on digital twins, comprising: Establish a 3D model of the SMT production line equipment and establish a mapping relationship between the virtual equipment status and the equipment joint parameters or key process parameters; Acquire the initial physical acquisition data of the SMT line equipment, perform time synchronization and integration based on process timing constraints, and output time-aligned line status data; Extract the overall line status data to construct multi-frame status data, perform singular value decomposition, and determine the state decomposition matrix and the proportion of each order component. Obtain the current rendering frame rate feedback result, and dynamically adjust based on the proportion of each order component and the current rendering frame rate feedback result to determine the number of retained components; The state decomposition matrix is ​​divided into primary and secondary components based on the number of retained components. The primary driving data is extracted from the primary components, and the secondary components are attenuated to generate secondary compensation data. The whole line rendering data is then fused and output. Calculate the image matching degree between the real image of the physical production line and the virtual image generated based on the whole line rendering data, and update the state decomposition matrix or execute rendering mode switching control according to the relationship between the image matching degree and the preset consistency threshold. Based on process sequence constraints, time synchronization and integration are performed, and the output time-aligned whole-line status data includes: Add a uniform absolute time stamp to the initial physical acquisition data, and resample according to the system simulation step size to generate single-frame synchronization state data; Obtain the exit trigger time of the upstream device and the inlet trigger time of the downstream device, as well as the device transmission parameters, and calculate the theoretical arrival time. Based on the comparison between the entry trigger time and the theoretical arrival time of the downstream equipment, the process timing constraint status is determined, and interpolation compensation is performed on the abnormal state quantities with timing conflicts. Perform singular value decomposition to determine the state decomposition matrix and the proportion of each order component, including: Perform mean-removal processing on multi-frame state data; Singular value decomposition is performed on the multi-frame state data after mean removal to extract the state decomposition matrix, singular value set, and time coefficient sequence that reflect the overall characteristics of the line. The proportion of each order component is then calculated based on the singular value set.

[0015] The specific implementation method of this embodiment includes: S101, establishing a three-dimensional model of the SMT production line equipment, and establishing a mapping relationship between the virtual equipment status and the equipment joint parameters or key process parameters; S102: Obtain the initial physical acquisition data of the SMT line equipment, perform time synchronization processing and data integration on the initial physical acquisition data based on the unified time stamp and process timing constraints, and output the line status data aligned by time. S103. Extract multiple single-frame synchronization state data in chronological order to construct continuous multi-frame state data, and perform singular value decomposition on the continuous multi-frame state data to determine the state decomposition matrix, singular value set, time coefficient sequence and proportion of each order component corresponding to the current time period. S104: Obtain the current rendering frame rate feedback result, compare and dynamically adjust the threshold based on the proportion of each order component and the current rendering frame rate feedback result, and determine the number of retained components currently participating in high-precision rendering. S105: Divide the state decomposition matrix into primary and secondary components according to the number of retained components. Perform inverse calculation on the primary components to extract the primary driving data, and perform attenuation processing on the secondary components to generate secondary compensation data. Combine the primary driving data and secondary compensation data to output the whole line rendering data. S106: Acquire real images of the physical production line and generate virtual images based on the 3D model and the whole line rendering data. Calculate the image matching degree between the real images of the physical production line and the virtual images. When the image matching degree is less than the preset consistency threshold, reconstruct state data for multiple consecutive frames and update the state decomposition matrix and singular value set. When the image matching degree is greater than or equal to the preset consistency threshold, execute the rendering mode switching control and output the 3D dynamic modeling simulation results. An SMT production line typically consists of solder paste printing equipment, inspection equipment, placement equipment, soldering equipment, and back-end inspection equipment connected in series. The sampling frequency, data format, and operation cycle of each piece of equipment are different. As production line cycle time increases, if static model overlay status display or simple event-triggered animation is still used, virtual actions may precede physical actions or key motion details may be missing, making it difficult to use 3D simulation results as a basis for debugging, monitoring and anomaly analysis. This application embodiment combines unified time alignment, continuous state decomposition, component retention control and image consistency verification in sequence, so that the entire 3D model can synchronize the real production line operation status under limited computing resources. In this embodiment, a three-dimensional model of the SMT production line is established, and a mapping relationship is established between the virtual equipment status and the equipment joint parameters or key process parameters. This includes establishing a driveable three-dimensional assembly model for various types of equipment in the production line, and mapping the motion nodes, posture nodes or process display nodes in the virtual model to the equipment data fields one by one. The 3D model includes, but is not limited to, the conveying mechanism model, the placement head model, the lifting mechanism model, the detection mechanism model, the furnace body conveying section model, and the printed circuit board model; The mapping relationship is achieved through a preset transfer function and matrix operations, where joint angles and linear displacements are determined by constructing... The homogeneous transformation matrix directly drives the local coordinate system of the skeletal nodes in the 3D model; temperature parameters The mapping to color display parameters uses a linear interpolation model: in, To calculate the target red-green-blue color vector for output, The current temperature parameter, and For the extreme values ​​of the range, and The base red, green, and blue color vectors and the maximum color change; For devices with multiple degrees of freedom, a multi-joint parameter table can be established according to the device's kinematic structure, and the device identifier, data field identifier, and rendering driver field can be used as associated indexes so that the corresponding virtual driving quantities can be calculated in reverse based on the state decomposition results. By pre-mapping equipment joint parameters or key process parameters to driveable nodes of a 3D model, the problem that subsequent decomposition results are difficult to use directly for 3D rendering can be solved. In this embodiment, the initial physical acquisition data of the SMT line equipment is acquired, and the initial physical acquisition data is processed for time synchronization and data integration based on a unified time stamp and process timing constraints. The line status data is output in time alignment. Specifically, the multi-source heterogeneous data collected from different equipment sides is organized into a unified and comparable status sequence. Initial physical acquisition data includes, but is not limited to, encoder pulses, position feedback, inlet sensor trigger information, outlet sensor trigger information, temperature sampling values, equipment operation flags, fault flags, and production line cycle time related data; a unified time stamp can be directly obtained from equipment that supports a unified time base, or a unified absolute time stamp can be added to the data outlet of equipment that does not support a unified time base. When integrating data, it is possible to first sort the data based on a unified time stamp, and then resample it according to the system simulation step size, so that data of different frequencies are organized into the same frame state section. By first unifying the time stamp, and then resampling and verifying the process constraints, the causal misalignment caused by the difference in sampling rhythm of data from multiple devices can be reduced. In this embodiment, multiple single-frame synchronization state data are extracted continuously in chronological order to construct continuous multi-frame state data. Singular value decomposition is performed on the continuous multi-frame state data to determine the state decomposition matrix, singular value set, time coefficient sequence and proportion of each order component corresponding to the current time period. Specifically, the state data under a unified time base is further converted into a mathematical expression that can be used for component retention judgment. Single-frame synchronization state data can be represented as a state vector containing the state quantities of multiple devices on the entire line. Multiple single-frame synchronization state data are stacked in chronological order to form continuous multi-frame state data. To avoid the influence of DC bias on the decomposition results, the continuous multi-frame state data is first processed to remove the mean and singular value decomposition is performed. The state decomposition matrix obtained after decomposition is used to describe the main spatial correlation features in the overall state change. The singular value set is used to represent the contribution of each order component. The time coefficient sequence is used to represent the change trend of each order component in continuous frames. The proportion of each order component can be calculated based on the proportion of the square quantity corresponding to the singular value set in the total amount. By decomposing the high-dimensional state of the entire line into several orders of components with different contribution sizes, a basis can be provided for determining the number of components to be retained in the future. Based on a unified timestamp and process sequence constraints, the initial physical acquisition data is processed for time synchronization and data integration, outputting time-aligned whole-line status data, including: S201 adds a unified absolute time stamp to the initial physical acquisition data of different SMT devices; S202, construct a circular buffer and arrange the initial physical acquisition data with uniform absolute time stamps in ascending order according to time sequence; S203, Resample and integrate the initial physical acquisition data after arrangement according to the system simulation step size to generate single-frame synchronization status data for the entire line; S204, Extract the upstream device exit trigger time and the downstream device entry trigger time for adjacent upstream and downstream devices in the SMT line equipment. S205: Obtain the preset equipment spacing, conveyor belt speed and conveying error, and determine the theoretical arrival time based on the equipment spacing, conveyor belt speed and conveying error; S206, compare the downstream equipment entry trigger time with the theoretical arrival time. If the downstream equipment entry trigger time is greater than or equal to the theoretical arrival time, it is determined that the process timing constraint is met, and the current single frame synchronization status data is retained. S207, if the downstream device entry trigger time is less than the theoretical arrival time, it is determined that there is a timing conflict. Based on the previous valid single frame synchronization state data and the next valid single frame synchronization state data at the current conflict time, the abnormal state quantity in the current single frame synchronization state data is interpolated and compensated. All processed single frame synchronization state data are summarized and the time-aligned whole line state data is output. In this embodiment, a unified absolute time stamp is added to the initial physical acquisition data of different SMT devices. Specifically, the data packets that were originally scattered under the local clocks of different devices are converted to a unified time reference. For devices that support a unified clock, their unified time field can be read directly. For devices that do not support a unified clock, a time stamp injection unit can be set on its data output side. By combining the pre-calibrated extreme value of the internal scanning cycle of the device with the communication link delay, the injected time stamp is back-calculated to compensate for communication delay, so that each frame of data is appended with a unified absolute time stamp restored to the physical sampling time. A data packet can be a single status record or a group of records collected from the same device at the same time. By providing the same comparison benchmark for data from each device, sorting distortion caused by independent clock drift can be reduced. In this embodiment, a circular buffer is constructed to arrange the initial physical acquisition data with a unified absolute time stamp in ascending order according to time sequence. Specifically, a time-series buffer space with buffer depth is established on the system side to temporarily store data packets from each device and rearrange them according to a unified time stamp. A circular buffer can continuously cover data from the most recent system cycles, thereby eliminating the adverse effects of network transmission jitter caused by inconsistent arrival times; the data in the buffer can be organized by dual indexing of device identifier and time stamp, and then written into a unified time base state stream when the output conditions are met; by setting a circular buffer and arranging it in ascending order, the impact of out-of-order arrival on the construction of subsequent single-frame synchronization states can be reduced. In this embodiment, the initial physical acquisition data after arrangement is resampled and integrated according to the system simulation step size to generate single-frame synchronization state data for the entire line. Specifically, the configuration is to project high-frequency data, low-frequency data and discrete event data into the same state frame based on a unified system cycle. For high-frequency continuous quantities, neighboring sampled values ​​or interval resampling can be used; for low-frequency field quantities, the most recent valid value can be retained using a short-time hold method; for event quantities, the trigger flag can be written into the status bit of the corresponding system cycle; the generated single-frame synchronization status data can include the displacement, velocity, status flag bits, and process parameter fields of multiple devices. By resampling and integrating the simulation step size of the unified system, the state data of each subsequent frame can have a consistent data structure. In this embodiment, for adjacent upstream and downstream devices in the SMT production line, the exit trigger time of the upstream device and the entry trigger time of the downstream device are extracted, and the theoretical arrival time is determined based on the device spacing, conveyor belt speed and conveying error. Specifically, the process flow relationship is introduced into the time verification process. The baseline value for the theoretical arrival time is: in, For theoretical arrival time, For the upstream equipment export trigger time, The preset equipment spacing, For conveyor belt speed; combined with conveyor error time The final theoretical allowable range for arrival time is: ; When the downstream equipment entry trigger time is earlier than the theoretical arrival time, it indicates that there is an abnormal process sequence relationship in the current data segment. By introducing process transfer logic between the upstream exit and the downstream entry, we can avoid retaining state frames that do not conform to the actual flow relationship based solely on time sorting. In this embodiment, if a timing conflict is determined, the abnormal state quantities in the current single-frame synchronization state data are interpolated and compensated based on the previous valid single-frame synchronization state data and the next valid single-frame synchronization state data at the current conflict time. Specifically, the state quantities in the conflict frame that are related to the causal contradiction of the process are corrected, rather than the entire frame data is directly deleted. Interpolation compensation can be achieved using linear interpolation, or spline interpolation when the state change is continuous, to obtain a smoother state transition. For continuous state variables, the rate of change is calculated as follows: in, The rate of change of the state variable. and These are the state variables corresponding to the next valid frame and the previous valid frame, respectively. and For the corresponding timestamp; if , To preset a stable judgment threshold, linear interpolation is used between the first and last points; like Furthermore, if the state variable is identified as a spatially continuous physical quantity, such as coordinate displacement or joint angle, then at least four consecutive valid frame data before and after the current conflict moment are extracted, and a cubic spline interpolation function is constructed to solve for the smooth compensation value at the current conflict moment; if the state variable is a discrete event trigger quantity, then interpolation is prohibited, and the Boolean state of the previous valid frame is forcibly used. For discrete state flags that are not suitable for interpolation, the previous value is preserved by default to maintain state stability and complete the correction. By interpolating and compensating abnormal state quantities, the rendering jumps caused by timing conflicts can be reduced while maintaining the continuity of the state flow. Multiple single-frame synchronization state data are extracted sequentially in time to construct continuous multi-frame state data. Singular value decomposition (SVD) is then performed on this continuous multi-frame state data to determine the state decomposition matrix, singular value set, time coefficient sequence, and the proportion of each order component for the current time period, including: S301, extract multiple consecutive single-frame synchronization state data from the time-aligned whole-line state data; S302, combine the multiple consecutive single-frame synchronization state data to construct multiple consecutive frame state data; S303, perform mean-removing processing on the multiple consecutive frame state data to generate a mean-removed data matrix. S304. Perform singular value decomposition on the mean-removed data matrix to extract the state decomposition matrix reflecting the overall state distribution characteristics, the singular value set reflecting the contribution of each component, and the time coefficient sequence describing the time change trend; S305. Determine the proportion of each order component based on the singular value set. In this embodiment, multiple consecutive single-frame synchronization state data are extracted from the time-aligned whole-line state data and combined to construct consecutive multi-frame state data. Specifically, a sliding window method is used to extract the latest consecutive frames from the unified state stream. Frame synchronization status data, and will contain The state dimension of the motion degrees of freedom of the equipment and key process fields within the entire production line is set as the rows of a matrix, and the time dimension, representing the state evolution over a continuous system cycle, is set as the columns of the matrix, constructing a matrix with a size of [missing information]. The state data matrix; Frames within a window The decomposition stability and system responsiveness can be set so that it can reflect the current operating characteristics without causing the decomposition results to be too lagging. By using multiple consecutive frames to form a two-dimensional matrix instead of single-frame data to participate in the decomposition, a complete mathematical basis can be provided for the subsequent extraction of the main common features in the linkage changes of the equipment. In this embodiment, mean-reduced processing is performed on multiple consecutive frames of state data to generate a mean-reduced data matrix. Specifically, this is configured to proceed along... The time dimension of the state data matrix is ​​the column direction. For each state variable row vector, it is based on the latest state within the current sliding window. Calculate the average time value and subtract it; This process does not incorporate future frame data after the current timestamp for averaging to meet the real-time physical causality constraints of the system; it highlights the dynamic changes relative to the average state by stripping the mean bias components of each state dimension. The data matrix obtained after mean removal is more suitable for characterizing the correlation between motion fluctuations, posture changes, and process quantity changes; by removing the mean without leading information, the interference of static bias on the singular value decomposition results can be reduced while maintaining temporal causality. In this embodiment, singular value decomposition is performed on the mean-removed data matrix to extract the state decomposition matrix, singular value set, and time coefficient sequence. The proportion of each order component is determined based on the singular value set. Specifically, the whole line change within the window is represented as the superposition of several order components. The proportion of each order component can be obtained by dividing the square of the corresponding singular value by the sum of the squares of all singular values, which is used to represent the degree of contribution of the component of that order to the overall change within the current window; components with a cumulative contribution proportion greater than or equal to a set threshold usually reflect the main action trend of multiple devices participating together, while components with a cumulative contribution proportion less than a set threshold usually correspond to local minor changes or higher-order disturbances. By calculating the set of singular values ​​and their proportions, a quantitative basis can be provided for the subsequent high-precision division of retained and attenuated components.

[0016] The number of components to be retained is dynamically adjusted based on the proportion of each component and the feedback results of the current rendering frame rate, including: The number of candidate truncated components is determined based on the proportion of each order of components and the preset weight threshold. If the current rendering frame rate feedback result is less than the system target rendering frame rate, then the preset weight threshold and the number of candidate truncated components are reduced; if the current rendering frame rate feedback result is greater than the sum of the system target rendering frame rate and the allowed upward difference, then the preset weight threshold and the number of candidate truncated components are increased; if the current rendering frame rate feedback result is greater than or equal to the system target rendering frame rate and less than or equal to the sum of the system target rendering frame rate and the allowed upward difference, then the number of candidate truncated components remains unchanged. Extracting primary driving data from the primary components and performing attenuation processing on secondary components to generate secondary compensation data includes: For the main components, the main driving data mapped to the equipment joint parameters or key process parameters are generated by reverse calculation. For the minor component, the characteristic coefficient value corresponding to the minor component is obtained from the time coefficient sequence generated by the singular value decomposition of multi-frame state data. Attenuation calculation is performed based on the time interval between the current time and the most recent update time of the characteristic coefficient value and the preset damping attenuation coefficient to generate minor compensation data.

[0017] In this embodiment, the current rendering frame rate feedback result is obtained, and threshold comparison and dynamic adjustment are performed based on the proportion of each order component and the current rendering frame rate feedback result to determine the number of retained components currently participating in high-precision rendering. Specifically, after obtaining the decomposition result, the number of retained components is adjusted in combination with the frame rate fed back by the rendering end. The current rendering frame rate feedback result can be returned by the rendering terminal cycle. The result can be the real-time frame rate of a single system cycle or the statistical frame rate over several system cycles. The proportion of each component reflects the importance of different components within the current window. The rendering frame rate feedback result reflects whether the current computing resources can support more components to participate in high-precision rendering. By combining the component percentage with the rendering frame rate feedback results for component selection, both rendering smoothness and the fidelity of the main actions can be balanced. In this embodiment, the state decomposition matrix is ​​divided into primary and secondary components according to the number of retained components. The primary components are inversely calculated to extract the primary driving data, and the secondary components are attenuated to generate secondary compensation data. The primary driving data and secondary compensation data are combined to output the full-line rendering data. Specifically, the components that contribute more are used for high-precision real-time driving, and the components that contribute less are maintained by attenuation approximation, thereby reducing the amount of computation. The primary driving data is used to drive the display of key mechanisms, key attitudes, and key processes, while the secondary compensation data is used to provide continuous compensation for weaker changes. The whole-line rendering data can be directly used as input to the 3D engine to drive the model skeleton, rigid body transformation, or process display layer updates. By distinguishing between primary and secondary components, the amount of high-frequency back calculation can be reduced without completely discarding weak component information. The method for determining the number of retained components currently participating in high-precision rendering by comparing and dynamically adjusting thresholds based on the proportion of each component and the current rendering frame rate feedback includes: S401, Obtain the preset weight threshold, and determine the number of candidate truncated components based on the proportion of each order component and the preset weight threshold; S402, Obtain the preset system target rendering frame rate and the allowable upward difference expressed in frames / second; S403, Compare the current rendering frame rate feedback result with the system target rendering frame rate and the sum of the system target rendering frame rate and the allowable upward difference. S404 If the current rendering frame rate feedback result is less than the system target rendering frame rate, then adjust the preset weight threshold according to the preset adjustment rules and redetermine the number of candidate truncated components in order to reduce the number of candidate truncated components. S405, if the current rendering frame rate feedback result is greater than the sum of the system target rendering frame rate and the allowable upward difference, then adjust the preset weight threshold according to the preset adjustment rules and redetermine the number of candidate truncated components to increase the number of candidate truncated components. S406, if the current rendering frame rate feedback result is greater than or equal to the system target rendering frame rate and less than or equal to the sum of the system target rendering frame rate and the allowable upward difference, then keep the current number of candidate truncated components unchanged and output the final number of retained components. In this embodiment, a preset weight threshold is obtained, and the number of candidate truncated components is determined based on the proportion of each level of components and the preset weight threshold. Specifically, the minimum number of components that meet the retention requirements is found according to the cumulative result of the component proportions. The preset weight threshold can be understood as the current cumulative contribution level that is allowed to be retained. When the components of each level are accumulated from high to low according to their proportions, the number of components corresponding to the first time the threshold is reached is the number of candidate truncated components. This method ensures that the number of candidate truncation components is not directly fixed, but automatically changes with the concentration of state changes within the current window; by calculating the number of candidate truncation components based on the proportion of each order component, components that have a greater impact on the current overall line change can be retained first. In this embodiment, the current rendering frame rate feedback result is compared with the system target rendering frame rate and its upward range. Specifically, the rendering performance requirements are incorporated into the adjustment process of the number of retained components. If the current rendering frame rate feedback result is lower than the system target rendering frame rate, it indicates that the current number of high-precision participating components exceeds the load threshold or the current rendering load exceeds the preset computing power limit. The screening intensity should be increased according to the preset adjustment rules to reduce the number of candidate truncated components. If the current rendering frame rate feedback result is higher than the sum of the system target rendering frame rate and the allowable upward difference, it means that the current computing resources meet the requirements. The screening intensity can be reduced according to the preset adjustment rules to allow more components to participate in high-precision rendering. If the current rendering frame rate is within the target range, the number of current candidate truncated components remains unchanged. The preset adjustment rules can be expressed as gradually increasing or decreasing the preset weight threshold, rather than changing it drastically all at once, in order to reduce frequent fluctuations in the number of retained components. A fixed adjustment step size is set. When the current rendering frame rate feedback result is less than the system target rendering frame rate, the preset weight threshold is subtracted from the adjustment step size. When the current rendering frame rate feedback result is greater than the sum of the system target rendering frame rate and the allowable upward difference, the preset weight threshold is added to the adjustment step size. At the same time, the value range of the preset weight threshold is set to 0.70 to 0.95. If the adjusted preset weight threshold exceeds this range, the boundary value is used as the current preset weight threshold. By introducing target frame rate comparison and threshold adjustment rules, the system can maintain a relatively stable rendering output under different computing loads. The method involves dividing the state decomposition matrix into primary and secondary components based on the number of retained components, performing inverse calculations on the primary components to extract primary driving data, and attenuating the secondary components to generate secondary compensation data. Finally, the method combines the primary driving data and the secondary compensation data to output the complete line rendering data. S501, decomposed components with serial numbers less than or equal to the number of retained components are marked as primary components, and decomposed components with serial numbers greater than the number of retained components are marked as secondary components; S502, for each primary component, extract the corresponding feature coefficients from the time coefficient sequence, perform inverse calculations on the feature coefficients and the corresponding decomposed columns in the state decomposition matrix, and generate primary driving data mapped to equipment joint parameters or key process parameters based on the mapping relationship; S503, for each secondary component, read the most recently updated feature coefficient value corresponding to the secondary component from the time coefficient sequence; S504, Obtain the time interval between the current time and the time corresponding to the most recently updated feature coefficient value; S505, Calculate the attenuation of the most recently updated feature coefficient value based on the time interval and the preset damping attenuation coefficient to generate secondary compensation data; S506, Combine and merge the main driving data and secondary compensation data to output the whole line rendering data. In this embodiment, decomposed components with an index less than or equal to the number of retained components are marked as primary components, and decomposed components with an index greater than the number of retained components are marked as secondary components. Specifically, the components are divided in descending order of singular value contribution. The primary components correspond to the overall linkage changes with higher contributions, which are suitable for high-frequency accurate back calculations; the secondary components correspond to local changes or weak disturbances with lower contributions, and no full high-frequency reconstruction is performed; by first completing the division of primary and secondary components, a foundation can be provided for subsequent processing paths with different computational intensities. In this embodiment, the corresponding feature coefficients are extracted for each major component, the feature coefficients are inversely calculated with the corresponding decomposition columns in the state decomposition matrix, and the main driving data is generated based on the mapping relationship. Specifically, the driving quantity of the component in the device parameter space is restored according to the component coefficients in the current window. In the specific inverse calculation, singular values ​​must be introduced as amplitude weights. The current frame feature coefficients extracted from the time coefficient sequence are multiplied by the corresponding singular values ​​in the singular value set, and then multiplied by the corresponding decomposition column vectors in the state decomposition matrix. This mathematically restores the dominant features after dimensionality reduction to the original high-dimensional physical state space. The inverse calculation result can be the mechanism displacement, joint angle, attitude change, or process quantity display value. For multi-device joint action scenarios, if the same main component has high weight in multiple state dimensions, it means that the component represents a type of whole-line coordinated action, which can drive the corresponding nodes of multiple device models simultaneously after inverse calculation. By performing a complete multiplication inverse calculation on the main component and mapping it to the device joint parameters or key process parameters, it can be ensured that the key actions are presented in the three-dimensional scene with high accuracy. In this embodiment, the most recently updated feature coefficient value is read for each minor component, and attenuation calculation is performed in combination with the time interval and the preset damping attenuation coefficient. Specifically, the minor component is no longer reconstructed in real time, but is approximated by its most recently valid value. The attenuation calculation can adopt an exponential attenuation method, so that the minor component gradually weakens with the system cycle, in order to maintain the continuity of the picture and reduce the sense of abrupt change. The decay calculation process is as follows: obtain the most recently updated feature coefficient values. And the time interval between the current time and the time corresponding to the most recent update. Combined with the preset damping attenuation coefficient The characteristic coefficient decay value of the current period is calculated using the exponential decay formula. in, is the base of the natural logarithm; Preset damping attenuation coefficient The calibration basis is as follows: obtain the natural decay characteristics of each minor component under historical normal operating conditions in the continuous system cycle, calculate the average time half-life when the amplitude drops to half of the initial value, and back-calculate the corresponding decay coefficient value. For example, if the most recent effective characteristic coefficient value of a certain component is 5.0, the preset damping attenuation coefficient is 0.1, and the computational resource allocation has not been updated frequently for two system cycles (i.e., the time interval is 2), then its currently used characteristic coefficient value is approximately calculated as follows: ; calculation yielded Afterwards, Multiply by the singular value corresponding to the minor component and the corresponding decomposition column vector in the state decomposition matrix, and map it back to the multidimensional device state space to generate secondary compensation data with consistent dimensions. This replaces the full high-frequency inverse calculation value, avoiding the problem of directly splicing the scalar coefficients with the high-dimensional state matrix due to dimensional mismatch; by applying damping attenuation approximation to minor components and reconstructing them in the same dimension, the continuous computational resource consumption of weak components can be reduced, while avoiding the instantaneous cut-off of minor details. In this embodiment, the main driving data and secondary compensation data are spliced ​​and fused to output the whole line rendering data. Specifically, in the unified rendering data structure, the main driving data vector and the secondary compensation data vector with the same device state dimension are directly added in the corresponding dimension to obtain the complete state vector after integrating the contributions of each order component. This vector is written as a high-precision driving field and a low-precision compensation field, and the rendering engine updates the 3D model according to the corresponding object level. After splicing and fusion, key mechanisms are directly controlled by the main driving data, while non-critical disturbances or gradual process changes are supplemented by secondary compensation data. By superimposing the main driving data and secondary compensation data into whole-line rendering data through same-dimensional matrix overlay, the realism of the main actions can be maintained while ensuring smooth transition of the screen, thus ensuring strict alignment of data fusion in the mathematical dimension.

[0018] Based on the relationship between image matching degree and preset consistency threshold, updating the state decomposition matrix or performing rendering mode switching control includes: If the image matching degree is less than the preset consistency threshold, the historical decomposition cumulative results are discarded, and multi-frame state data is reconstructed to update the state decomposition matrix. If the image matching degree is greater than or equal to the preset consistency threshold, the current state decomposition matrix is ​​maintained, and the rendering mode switching control is executed. The execution of rendering mode switching control includes: Extract the singular value set generated by singular value decomposition from multiple frames of state data, and calculate the ratio of singular value extrema in the current singular value set; If the singular value-to-extreme ratio is greater than the preset ratio threshold, the system will switch to simplified rendering and motion response mode; if the singular value-to-extreme ratio is less than or equal to the preset ratio threshold, the primary and secondary processing flow will be restored. Please see Figure 2 A digital twin-based 3D dynamic modeling and simulation system for an entire SMT production line includes: The model building module is used to build a 3D model of the SMT production line equipment and establish a mapping relationship between the virtual equipment status and the equipment joint parameters or key process parameters. The data acquisition and synchronization module is used to acquire initial physical data, synchronize and integrate it in time, and output time-aligned whole-line status data. The decomposition and analysis module is used to construct multi-frame state data and perform singular value decomposition. The rendering precision control module is used to dynamically determine the number of components to be retained based on the component ratio and rendering frame rate feedback results. The data merging module is used to process the primary and secondary components according to the number of retained components, and merge and output the whole line rendering data. The virtual-real image comparison module is used to calculate the image matching degree between real and virtual images, update the state decomposition matrix, or perform rendering mode switching control.

[0019] In this embodiment, a real image of the physical production line is acquired, and a virtual image is generated based on the 3D model and the whole line rendering data. The image matching degree between the real image of the physical production line and the virtual image is calculated. When the image matching degree is less than a preset consistency threshold, the state data of multiple consecutive frames is reconstructed and the state decomposition matrix and singular value set are updated. When the image matching degree is greater than or equal to the preset consistency threshold, the rendering mode switching control is executed. Specifically, an image comparison-based verification path is added in addition to the rendering result. The physical production line's real images are continuously acquired from fixed positions, while the virtual images are output as keyframes from the 3D model at the corresponding viewing angle. The two are compared within the same set period to determine whether the current state decomposition results are still applicable to the working conditions. To eliminate cross-domain imaging bias between the differences in ambient lighting in the physical production line and the ideal material representation in the virtual model, before comparing the images, the high-frequency edge structures of the real and virtual images of the physical production line are extracted using an edge detection operator to generate corresponding binarized edge feature maps. Based on edge feature maps, image matching degree can be evaluated by structural similarity calculation method, which evaluates the consistency of the overall structure and display changes of the image. The calculation method is as follows: divide the real image of the physical production line and the virtual image into multiple local image blocks of the same size, and calculate the mean brightness, variance and covariance of each corresponding local image block. For two corresponding local image patches and By extracting local image patches separately and average brightness and ,variance and and the covariance between the two It also introduces a pre-defined positive constant that approaches zero to prevent the denominator from being zero. and Substituting into the formula, its structural similarity index is calculated. : After calculating the structural similarity index of each local image patch pair, the mean of the structural similarity indices of all local image patches is used as the image matching degree. The basis for obtaining the preset consistency threshold is as follows: during the debugging phase when the equipment is running normally and there is no misalignment between virtual and real images, multiple sets of reference virtual and real image pairs are collected, the statistical average of the corresponding image matching degree is calculated, and the statistical average is set down by a certain percentage as the preset consistency threshold; by comparing virtual and real images, the mismatch of decomposition results caused by equipment replacement, long-term operation or changes in operating conditions can be detected in a timely manner. In this embodiment, the rendering mode switching control is executed, and the three-dimensional dynamic modeling simulation results are output. Specifically, when the image matching degree meets the requirements, the singular value spectrum features are used to determine whether it is appropriate to maintain the component-driven mode. If the singular value extreme value ratio indicates that the components are relatively dispersed, the real-time feature decomposition process is no longer maintained, and the simplified rendering and action response mode is switched to be executed according to the preset device event table. If the singularity-to-extreme ratio indicates that the current components are relatively concentrated, then the processing flow of the primary and secondary components is maintained or restored; the preset equipment event table can record the correspondence between basic events such as equipment start-up and shutdown, conveying trigger, arrival trigger, and process stage switching and their corresponding simplified actions; In the simplified rendering and motion response mode, the system stops performing singular value decomposition and subsequent principal and secondary component inverse calculation on continuous multi-frame state data. Instead, it extracts discrete event trigger flags from the time-aligned whole-line state data. When the corresponding event is detected, it directly calls the standard animation clip sequence bound in the preset device event table and completes the basic displacement animation update of the rigid body or joint of the model within a preset fixed time beat, thereby bypassing the matrix inverse calculation and feature mapping logic. By setting up two levels of control—image consistency judgment and mode switching judgment—simplified animation triggering methods can be switched in a timely manner to maintain stable system operation when the operating conditions change significantly or the components are discrete, causing the feature decomposition to be unable to stably extract the dominant features. A method for acquiring real images of the physical production line and generating virtual images based on 3D models and overall production line rendering data; calculating the image matching degree between the real physical production line images and the virtual images; reconstructing state data for multiple consecutive frames and updating the state decomposition matrix and singular value set when the image matching degree is less than a preset consistency threshold; and executing rendering mode switching control when the image matching degree is greater than or equal to the preset consistency threshold, including: S601, Based on the virtual camera configured in the 3D model, capture key frame images as virtual images according to a set period; S602, acquire the real images of the physical production line collected by fixed positions on the physical production line, compare the real images of the physical production line with the virtual images, and calculate the image matching degree. S603, obtain the preset consistency threshold, and compare the image matching degree with the preset consistency threshold; S604, if the image matching degree is less than the preset consistency threshold, determine that the current decomposition result does not match, discard the historical decomposition cumulative result, call the latest time-aligned whole line state data to reconstruct the continuous multi-frame state data, and update the state decomposition matrix and singular value set. S605, if the image matching degree is greater than or equal to the preset consistency threshold, then maintain the current state decomposition matrix and execute the rendering mode switching control; In this embodiment, based on the virtual camera configured in the 3D model, key frame images are captured as virtual images according to a set period. Specifically, the configuration is to preset the observation angle corresponding to the fixed camera position in the 3D scene and periodically output the current key rendering frame. The virtual camera's field of view, focal length range, and viewing direction can be kept as consistent as possible with the fixed camera position to improve the effectiveness of image comparison; key frames can be selected to include the main equipment action area and the board conveying area; by using a virtual camera corresponding to the fixed camera position to output virtual images, the consistency of virtual and real image comparison can be improved. In this embodiment, the real images of the physical production line are compared with the virtual images to calculate the image matching degree. Specifically, the configuration is to perform structural similarity calculation on the real and virtual images obtained within the same set period. Before comparison, bilinear interpolation algorithm can be used to unify the resolution of the two types of images, crop the viewing area according to the preset effective observation bounding box, and divide the channel gray value of all pixels by 255 to perform brightness normalization processing to reduce the impact of non-critical imaging differences on the matching degree. The lower the image matching degree, the greater the deviation between the current 3D model rendering result and the real production line scene; by periodically calculating the image matching degree, the 3D simulation results can be continuously checked against the real scene performance. In this embodiment, if the image matching degree is less than the preset consistency threshold, it is determined that the current decomposition result is mismatched, the historical decomposition cumulative result is discarded, the latest time-aligned whole line state data is called to reconstruct the continuous multi-frame state data, and the state decomposition matrix and singular value set are updated. Specifically, when it is confirmed that the original decomposition result can no longer correctly express the current working condition, the old decomposition basis is not continued to be used, but the decomposition establishment process is re-executed. The decomposition matrix and singular value set are reconstructed using the latest time-aligned whole-line state data as input, so that the new state decomposition matrix and singular value set match the current running state; by reconstructing the decomposition results when the image matching degree is insufficient, the mismatch effect caused by the long-term accumulation of old decomposition results can be reduced. Methods for controlling rendering mode switching and outputting 3D dynamic modeling and simulation results include: S701, Extract the minimum and maximum singular values ​​from the current singular value set; S702, Calculate the ratio of the minimum singular value to the maximum singular value to obtain the singular value extreme value ratio; S703, Obtain a preset ratio threshold and compare the singular value extreme value ratio with the preset ratio threshold; S704, Obtain a preset device event table; S705, if the singular value extreme value ratio is greater than the preset ratio threshold, it is determined that each component is scattered and lacks a dominant component, the real-time feature decomposition process is paused, and the simplified rendering and action response mode is switched to be executed according to the preset device event table, and the 3D dynamic modeling simulation results are output. S706: If the singular value extreme value ratio is less than or equal to the preset ratio threshold, the component is determined to be concentrated, the simplified rendering and motion response mode is exited, the primary and secondary processing flow of the main and secondary components is restored, and the 3D dynamic modeling simulation results are output. In this embodiment, the minimum and maximum singular values ​​in the current singular value set are extracted, and the ratio between the two is calculated to obtain the singular value extreme value ratio. Specifically, the concentration of the singular value spectrum is used to determine whether there is a significant dominant component in the current window. If the difference between the minimum singular value and the maximum singular value is less than the preset distribution threshold, it indicates that the dispersion of each component contribution is greater than the preset dispersion threshold, and the current decomposition result is difficult to be stably represented by a small number of main components; conversely, if the ratio is less than the preset ratio threshold, it indicates that the concentration of the dominant component is greater than the preset concentration threshold, and it is suitable to adopt the component retention and primary and secondary component processing method. The calibration method for the preset ratio threshold is as follows: extract continuous multi-frame state data of historical normal operation and containing clear dominant actions, perform singular value decomposition and calculate the distribution range of the corresponding singular value extreme value ratio, take the upper limit of the distribution range as the preset ratio threshold, so as to quantitatively determine the concentration of the current state components; by calculating the singular value extreme value ratio, a simple criterion can be provided for whether to continue to use the component-driven mode. In this embodiment, a preset device event table is obtained, and when the singular value extreme value ratio is greater than a preset ratio threshold, the simplified rendering and action response mode executed according to the preset device event table is switched. Specifically, when the components in the current window lack a clear primary and secondary relationship, the real-time feature decomposition process is no longer executed, but instead, the rendering mode is returned to the basic event-driven rendering mode. The device event table can be pre-configured with the correspondence between device operation events, transfer start events, transfer arrival events, detection completion events, and process switching events and simplified actions, so that the 3D scene can at least maintain the response of major events; by switching to simplified rendering and action response mode when components are dispersed, the computational pressure caused by continuing high-frequency decomposition under unstable working conditions can be alleviated, and display instability problems can be avoided. In this embodiment, if the singular value extreme value ratio is less than or equal to a preset ratio threshold, it is determined that the components are concentrated, the simplified rendering and action response mode is exited, and the primary and secondary processing flow of the main components and secondary components is restored. Specifically, when the subsequent state window has a clearer dominant component structure again, the system re-uses the high-precision driving of the main components and the attenuation compensation of the secondary components to generate the whole line rendering data. By switching between two rendering modes based on the singularity extremum ratio, the results of 3D dynamic modeling and simulation can be made more adaptable to different running stages. The model building module is used to build a three-dimensional model of the SMT production line equipment and establish a mapping relationship between the virtual equipment status and the equipment joint parameters or key process parameters. The data acquisition and synchronization module is used to acquire the initial physical acquisition data of the SMT line equipment. Based on the unified time stamp and process timing constraints, the initial physical acquisition data is processed for time synchronization and data integration, and outputs the line status data aligned by time. The line status data includes multiple single-frame synchronized status data arranged in time order. The decomposition and analysis module is used to extract multiple single-frame synchronization state data in chronological order to construct continuous multi-frame state data, and to perform singular value decomposition on the continuous multi-frame state data to determine the state decomposition matrix, singular value set, time coefficient sequence and proportion of each order component for the current time period. The rendering precision control module is used to obtain the current rendering frame rate feedback result, compare and dynamically adjust the threshold based on the proportion of each order component and the current rendering frame rate feedback result, and determine the number of retained components currently participating in high-precision rendering. The driving data merging module is used to divide the state decomposition matrix into primary and secondary components according to the number of retained components, perform inverse calculation on the primary components to extract the primary driving data, and perform attenuation processing on the secondary components to generate secondary compensation data. The primary driving data and secondary compensation data are combined to output the whole line rendering data. The virtual-real image comparison module is used to acquire real images of the physical production line and generate virtual images based on the 3D model and the whole line rendering data. It calculates the image matching degree between the real images of the physical production line and the virtual images. When the image matching degree is less than the preset consistency threshold, it reconstructs the state data of multiple consecutive frames and updates the state decomposition matrix and singular value set. When the image matching degree is greater than or equal to the preset consistency threshold, it executes the rendering mode switching control and outputs the 3D dynamic modeling simulation results. In this embodiment, the model building module receives the model information and mapping configuration relationship of the entire production line equipment and parses it into a driveable three-dimensional object structure, which can improve the correspondence efficiency between subsequent rendering data and model nodes; the data acquisition and synchronization module processes multi-source heterogeneous data through unified time stamps, circular buffers and process timing constraints, which improves the time consistency of the entire production line status data. The decomposition and analysis module improves the ability to extract the main linkage changes of the entire line by performing singular value decomposition on state data of multiple consecutive frames; the rendering precision control module improves the adaptability of the number of high-precision rendering components by comparing the component ratio with the rendering frame rate feedback results. The driving data merging module improves rendering continuity under limited computing resources by back-calculating the main components and attenuating the secondary components; the virtual-real image comparison module improves the consistency and operational stability between the 3D dynamic modeling simulation results and the real production line status by judging the image matching degree and controlling the mode switching.

[0020] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A three-dimensional dynamic modeling and simulation method for an entire SMT production line based on digital twins, characterized in that, include: Establish a 3D model of the SMT production line equipment and establish a mapping relationship between the virtual equipment status and the equipment joint parameters or key process parameters; Acquire the initial physical acquisition data of the SMT line equipment, perform time synchronization and integration based on process timing constraints, and output time-aligned line status data; Extract the overall line state data to construct multi-frame state data, perform singular value decomposition, and determine the state decomposition matrix and the proportion of each order component. Obtain the current rendering frame rate feedback result, and dynamically adjust based on the proportion of each order component and the current rendering frame rate feedback result to determine the number of retained components; The state decomposition matrix is ​​divided into primary and secondary components according to the number of retained components. The primary driving data is extracted from the primary components, and the secondary components are attenuated to generate secondary compensation data. The whole line rendering data is then fused and output. Calculate the image matching degree between the real image of the physical production line and the virtual image generated based on the whole line rendering data, and update the state decomposition matrix or execute rendering mode switching control according to the relationship between the image matching degree and the preset consistency threshold.

2. The SMT whole-line three-dimensional dynamic modeling and simulation method based on digital twin according to claim 1, characterized in that, Based on process sequence constraints, time synchronization and integration are performed, and the output time-aligned whole-line status data includes: Add a uniform absolute time stamp to the initial physical acquisition data, and resample according to the system simulation step size to generate single-frame synchronization state data; Obtain the exit trigger time of the upstream device and the inlet trigger time of the downstream device, as well as the device transmission parameters, and calculate the theoretical arrival time. Based on the comparison between the entry trigger time of the downstream equipment and the theoretical arrival time, the process timing constraint status is determined, and interpolation compensation is performed on abnormal state quantities with timing conflicts.

3. The SMT whole-line three-dimensional dynamic modeling and simulation method based on digital twin according to claim 1, characterized in that, Perform singular value decomposition to determine the state decomposition matrix and the proportion of each order component, including: The mean value of the multi-frame state data is removed. Singular value decomposition is performed on the multi-frame state data after the mean removal process to extract the state decomposition matrix, singular value set, and time coefficient sequence that reflect the overall characteristics of the line. The proportion of each order component is calculated based on the singular value set.

4. The SMT whole-line three-dimensional dynamic modeling and simulation method based on digital twin according to claim 1, characterized in that, The number of retained components is determined by dynamically adjusting based on the proportion of each component and the current rendering frame rate feedback result, including: The number of candidate truncated components is determined based on the proportion of each order of components and the preset weight threshold. If the current rendering frame rate feedback result is less than the system target rendering frame rate, then the preset weight threshold and the number of candidate truncated components are reduced; if the current rendering frame rate feedback result is greater than the sum of the system target rendering frame rate and the allowed upward difference, then the preset weight threshold and the number of candidate truncated components are increased; if the current rendering frame rate feedback result is greater than or equal to the system target rendering frame rate and less than or equal to the sum of the system target rendering frame rate and the allowed upward difference, then the number of candidate truncated components remains unchanged.

5. The SMT whole-line three-dimensional dynamic modeling and simulation method based on digital twin according to claim 1, characterized in that, Extracting primary driving data from the primary components and performing attenuation processing on secondary components to generate secondary compensation data includes: For the main components, the main driving data mapped to the joint parameters or key process parameters of the equipment are generated by reverse calculation; For the minor component, the characteristic coefficient value corresponding to the minor component is obtained from the time coefficient sequence generated by the singular value decomposition of multi-frame state data. Attenuation calculation is performed based on the time interval between the current time and the most recent update time of the characteristic coefficient value and the preset damping attenuation coefficient to generate the minor compensation data.

6. The SMT whole-line three-dimensional dynamic modeling and simulation method based on digital twin according to claim 1, characterized in that, Updating the state decomposition matrix or performing rendering mode switching control based on the relationship between the image matching degree and the preset consistency threshold includes: If the image matching degree is less than the preset consistency threshold, the historical decomposition cumulative results are discarded, and multi-frame state data is reconstructed to update the state decomposition matrix. If the image matching degree is greater than or equal to the preset consistency threshold, the current state decomposition matrix is ​​maintained, and rendering mode switching control is executed.

7. The SMT whole-line three-dimensional dynamic modeling and simulation method based on digital twin according to claim 6, characterized in that, The execution of rendering mode switching control includes: Extract the singular value set generated by singular value decomposition of the multi-frame state data, and calculate the ratio of singular value extrema in the current singular value set; If the singular value extreme value ratio is greater than a preset ratio threshold, it is determined to switch to simplified rendering and action response mode; if the singular value extreme value ratio is less than or equal to the preset ratio threshold, the primary and secondary processing flow is restored.

8. A three-dimensional dynamic modeling and simulation system for an entire SMT production line based on digital twins, implementing the method described in any one of claims 1-7, characterized in that, include: The model building module is used to build a 3D model of the SMT production line equipment and establish a mapping relationship between the virtual equipment status and the equipment joint parameters or key process parameters. The data acquisition and synchronization module is used to acquire initial physical data, synchronize and integrate it in time, and output time-aligned whole-line status data. The decomposition and analysis module is used to construct multi-frame state data and perform singular value decomposition. The rendering precision control module is used to dynamically determine the number of components to be retained based on the component ratio and rendering frame rate feedback results. The data merging module is used to process the primary and secondary components according to the number of retained components, and merge and output the whole line rendering data. The virtual-real image comparison module is used to calculate the image matching degree between real and virtual images, update the state decomposition matrix, or perform rendering mode switching control.

Images