An injection machine electro-hydraulic coupling semi-physical simulation platform and control system

Abstract

This invention discloses a semi-physical simulation platform and control system for a dispensing machine, comprising: a mechanical execution module: simulating nozzle deformation through a quantum dispensing head virtual environment to generate first physical data, and integrating a rigid-flexible coupling platform and a laser interferometer to capture high-speed trajectory pose deviations to generate second physical data; a digital hydraulic drive module: simulating fault repair dynamics using a converter pump and a quantum attenuator to generate third physical data; an edge-cloud collaboration module: constructing a twin based on an FPGA-quantum architecture to generate fourth physical data, and generating fifth physical data through transfer learning AI algorithms to optimize hydraulic pump and valve strategies; a global monitoring and diagnostic module: constructing a multi-physical quantity model through fiber optic-quantum sensors to generate seventh physical data to evaluate wide-temperature-range stability, and generating sixth physical data; and a simulation comprehensive management module: evaluating the platform's stability under wide-temperature-range and complex vibration scenarios.

Description

Technical Field

[0001] This invention relates to the field of hardware-in-the-loop simulation technology, and in particular to a hardware-in-the-loop simulation platform and control system for a dispensing machine with electromechanical-hydraulic coupling. Background Technology

[0002] As a key supporting equipment in modern precision manufacturing, dispensing machines play an indispensable role in many fields such as electronic packaging, automobile manufacturing, and medical devices. Driven by both consumption and industrial upgrading, the market has put forward higher requirements for dispensing processes—it must not only meet the mass production needs of high precision and high efficiency, but also adapt to the challenges of diverse materials and complex working conditions. This has prompted dispensing machines to evolve from single-function equipment to intelligent and integrated ones, becoming an important cornerstone for promoting the high-quality development of precision manufacturing.

[0003] However, traditional simulation methods are often based on idealized models or static parameter designs, making it difficult to accurately capture the dynamic characteristics of actual hardware during high-speed, high-precision operations. This leads to a significant mismatch between simulation results and actual hardware performance. Traditional simulations often neglect the dynamic coupling effect of mechanical, hydraulic, and thermal multi-physics fields, resulting in significant deviations between the actual dispensing trajectory of the nozzle and the simulation prediction. For example, when a certain type of dispensing machine operates at a high-speed curved trajectory of 300 mm / s, the traditional simulation predicts that the nozzle pose deviation is less than 50 μm. However, actual measurements show that mechanical vibration causes radial movement of the nozzle to reach 80-120 μm, hydraulic fluctuations cause pressure pulsation of ±0.2 MPa, and temperature changes cause viscosity fluctuations of ±15%. Ultimately, the actual dispensing trajectory deviation reaches 150-200 μm, exceeding the allowable range of the process, resulting in excessive chip bonding voids and reduced product yield. Therefore, a semi-physical simulation platform and control system for electromechanical-hydraulic coupling of a dispensing machine is proposed. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a semi-physical simulation platform and control system for a glue dispensing machine with electromechanical-hydraulic coupling.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A semi-physical simulation platform and control system for an electromechanical-hydraulic coupling dispensing machine, comprising: Mechanical execution module: Constructs a virtual environment for the quantum dispensing head, simulates nozzle deformation, generates first physical data on glue content distribution and response delay, inputs it into the digital twin module to calibrate physical model parameters, indirectly improving the simulation fidelity of the fourth data; integrates a rigid-flexible coupling platform and a laser interferometer to capture high-speed trajectory pose deviation and mechanical clearance characteristics, generates second physical data, and transmits it in real time to the global monitoring and diagnosis module to adjust the model's feature weights and thresholds, optimizing the sixth physical data; Digital hydraulic drive module: Simulates fault and maintenance dynamic response through converter pump and quantum attenuator, generates third physical data including pressure fluctuation suppression and energy efficiency recovery rate, inputs to the global monitoring module to adjust the generation logic of sixth physical data, and synchronizes the parameters to the digital twin module to drive mechanical execution; Edge-cloud collaborative digital twin module: Based on FPGA-quantum architecture, a twin is built to generate fourth physical data containing instruction delay and simulation error. The accuracy of the fault prediction model of the sixth physical data is adjusted. Transfer learning AI algorithm is deployed to simulate the adjustment of new material parameters. Fifth physical data containing model generalization and accuracy improvement rate is generated. The hydraulic control parameter generation logic of the third physical data is adjusted. Instructions are sent to the mechanical execution module. The full-domain monitoring and diagnosis module constructs a multi-physical quantity model through fiber optic-quantum sensors, simulates environmental interference, generates the seventh physical data to evaluate the stability over a wide temperature range, simulates faults and maintenance operations, generates the sixth physical data containing fault identification and maintenance efficiency improvement rates, and adjusts the AI ​​algorithm feature weights of the fifth physical data. Simulation Integrated Management Module: Evaluates the stability of the platform under wide temperature range and complex vibration scenarios. Based on the stability evaluation results, it generates engineering parameter optimization suggestions and feeds them back to the mechanical actuation and digital hydraulic drive modules.

[0006] The above technical solution further includes: Furthermore, the mechanical actuation module includes: Infrastructure Simulation Unit: A virtual testing environment is constructed based on a quantum-enhanced piezoelectric metamaterial dispensing head. A topology optimization model is used to accurately simulate the dynamic deformation process of the nozzle at the sub-micron scale. First, utilizing the piezoelectric effect and quantum tunneling characteristics of the quantum-enhanced piezoelectric metamaterial, a high-precision constitutive model of the nozzle material is established. The design domain is defined as the key deformation regions of the nozzle (such as the outlet section and support structure). Constraints include material stiffness threshold, maximum allowable displacement deviation, and the required glue distribution range. Subsequently, a density-based topology optimization algorithm is used to iteratively optimize the material distribution and structural topology to minimize dynamic deformation error, generating a nozzle geometric model that meets sub-micron precision requirements. This is then simulated using a finite element method (FEM) to simulate the final model. Dispersive element coupling simulation simulates the dynamic response of a nozzle under high-speed motion when subjected to the coupling effects of multiple physical fields such as hydraulic shock and mechanical vibration. It calculates the displacement field, strain field, and colloid flow characteristics at various points on the nozzle surface in real time. Finally, it extracts the colloid volume distribution (spatial distribution of colloid volume flow rate based on fluid dynamics simulation) and response delay (time difference from the issuance of control command to the actual deformation of nozzle) data to form the first physical simulation data containing colloid volume distribution and response delay. This data is directly input into the edge-cloud collaborative digital twin module. The nonlinear characteristic parameters of the nozzle stiffness matrix and driving voltage-displacement response curve in the digital twin are calibrated through parameter mapping algorithm to ensure the consistency between the virtual environment and the actual hardware in submicron-level dynamic characteristics. The physical equipment simulation unit integrates a rigid-flexible coupling motion platform and a laser interferometer to simulate the mechanical execution process of a dispensing head on a high-speed curved trajectory and capture its physical operation characteristics. First, the rigid-flexible coupling motion platform achieves macroscopic trajectory motion through a rigid body module, while the flexible module simulates the minute elastic deformation of the dispensing head support. The two work together to reproduce the motion characteristics under actual working conditions. The laser interferometer measures the motion platform's pose (X / Y / Z axis displacement and rotation angle) and the spatial trajectory of the dispensing head tip in real time, simultaneously acquiring mechanical clearance data (such as joint bearing wear and transmission chain elastic deformation). The measurement signals are processed to calculate the motion platform's pose deviation (root mean square error between the actual and ideal trajectories) and the accumulated mechanical clearance (the amount of mechanical clearance). The statistical characteristics of clearance changes during motion are analyzed, and the force transmission efficiency (the ratio of output force to input driving force, reflecting energy transmission loss) is calculated by combining force sensor data. Finally, a second physical simulation data containing trajectory error and force transmission efficiency is generated. This data is transmitted in real time to the multi-physical quantity fusion global monitoring module via the TSN network as a feature input for the fault prediction model (such as triggering an abnormal mechanical clearance warning when the trajectory error exceeds the threshold). At the same time, it is fed back to the digital twin module, which updates the mass-damping-stiffness parameters of the motion platform through the dynamic model correction algorithm, and optimizes the trajectory planning accuracy of the control commands (such as using the B-spline curve smoothing algorithm to reduce trajectory jitter), forming a closed-loop control link of simulation-measurement-feedback-optimization.

[0007] Furthermore, the digital hydraulic drive module includes: Hydraulic Fault Dynamic Response Simulation Unit: Through the synergistic effect of a variable flow hydraulic pump and a quantum piezoelectric digital pulsation attenuator, it accurately simulates the fault state and maintenance operations of a hydraulic system. Typical fault modes (such as pump efficiency dropping to below 85% and valve leakage exceeding 0.1 mL / s) are actively injected into the virtual test environment. The pressure fluctuation signal of the hydraulic pipeline is collected in real time through the quantum piezoelectric digital pulsation attenuator. Combined with the flow-pressure characteristic curve of the variable flow pump, the dynamic response characteristics of the hydraulic system under fault conditions are quantified. Maintenance operations such as resetting simulated parameters (e.g., adjusting the gain of the PID controller) and structural adjustments (e.g., optimizing the valve opening) are carried out by adjusting the output flow and pressure of the hydraulic pump in real time through a closed-loop control algorithm and suppressing pressure fluctuations by utilizing the active damping control function of the pulsation attenuator. Based on real-time data collected by the pressure sensor array, the pressure fluctuation suppression effect (such as peak pressure reduction rate and fluctuation frequency attenuation rate) and energy efficiency recovery rate (the ratio of total system energy consumption after maintenance to energy consumption under fault conditions) are calculated, forming a third physical simulation data that includes the pressure fluctuation suppression effect and energy efficiency recovery rate. This data is input into the full-domain monitoring and diagnosis module to verify the effectiveness of the maintenance strategy and is fed back to the digital twin module to adjust the hydraulic impedance matching algorithm (such as optimizing the matching relationship between pipeline diameter, valve opening and pump displacement). Hydraulic-Mechanical-Thermal Dynamic Coupled Simulation Unit: The finite element method is used to establish a mechanical vibration sub-model (simulating the elastic deformation of the glue injection head bracket), a computational fluid dynamics simulation hydraulic fluctuation sub-model (analyzing the hydraulic oil flow characteristics), and a heat conduction sub-model (calculating temperature distribution and viscosity changes). The common node matching of the three domain meshes is achieved through mesh coupling technology to ensure the consistency of boundary conditions. During the glue injection process, the mechanical vibrations caused by hydraulic shock (such as the vibration of the glue injection head caused by sudden pressure changes when the pump starts and stops), the thermal effects caused by vibration (such as frictional heat generation changing the viscosity of hydraulic oil), and the influence of heat conduction on mechanical properties (such as the decrease in material stiffness caused by temperature rise) are simulated. The interaction of each physical field is calculated in real time through a multi-physics coupling solver to generate coupling effect analysis results (such as the influence coefficient of vibration frequency on glue injection accuracy and the influence curve of temperature gradient on hydraulic system efficiency). The coupling effect analysis results are input into the digital twin module to verify the robustness of the control algorithm under complex working conditions (such as anti-interference ability and trajectory tracking accuracy), and fed back to the mechanical execution module to optimize the glue injection trajectory planning (such as avoiding vibration-sensitive areas and adjusting the glue injection speed to reduce the impact of thermal effects).

[0008] Furthermore, the edge-cloud collaborative digital twin electronic control module includes: A digital twin is constructed based on an FPGA-quantum computing collaborative architecture. The FPGA, as a classical computing acceleration unit, is responsible for real-time processing of sensor data, execution of control algorithms and communication protocol stacks. It simulates complex nonlinear physical processes (such as high-order elastic deformation of nozzle materials and quantum tunneling effect of hydraulic systems) through qubits. The two achieve data interaction and task splitting through a high-speed bus. For example, the FPGA processes TSN network data packets synchronized at the sub-microsecond level, while the quantum computing module optimizes the nonlinear parameter calibration of the nozzle stiffness matrix in the digital twin. First, based on the physical properties of the quantum-enhanced piezoelectric metamaterial dispensing head, a quantum-state-described nozzle dynamics model is established in the quantum computing module. Subsequently, the FPGA is used to deploy the model in real time through a hardware description language, forming a hybrid computing architecture of quantum modeling and classical acceleration.

[0009] For high-dimensional physical models (such as finite element models with millions of degrees of freedom), a model reduction method based on POD (orthogonal eigenvalue decomposition) is adopted to extract the minority modes that dominate the dynamic characteristics (such as the first 10 vibration modes), thus compressing the model size to a scale that can be computed in real time (such as from 10). 6 The degree of freedom is reduced to 10³ degrees of freedom. At the same time, the nonlinear part is approximated by combining the Krylov subspace method. The reduced-order model is deployed through FPGA hardware accelerator and uses parallel computing units to achieve more than one million model updates per second, supporting the real-time simulation requirements in HIL testing.

[0010] During the operation of the digital twin, the FPGA collects the timestamps of the control command issuance and the actual hardware execution in real time, and calculates the control command delay (such as the time difference from the FPGA issuing the command to the dispensing head starting to move); at the same time, by comparing the nozzle displacement predicted by the digital twin with the displacement data measured by the laser interferometer, the simulation error is quantified (such as root mean square error ≤ 0.05μm). The generated fourth physical simulation data includes indicators such as control command delay and simulation error, which directly support the verification and virtual debugging of control strategies in HIL testing. For example, if the simulation error exceeds the threshold, the model correction process is triggered, and the nozzle stiffness parameters are recalibrated through the quantum computing module. At the same time, the data is input into the global monitoring module to evaluate the real-time performance of the control strategy (such as whether the command delay meets the ±1μs process requirement) and fed back to the mechanical execution module to optimize the dispensing head drive voltage curve (such as adjusting the PWM waveform parameters to reduce response delay).

[0011] Furthermore, the edge-cloud collaborative digital twin electronic control module also includes: Based on pre-trained deep learning models (such as convolutional neural networks + long short-term memory networks), an integrated transfer learning module is used to transfer knowledge across working conditions. The pre-trained model completes initial training on working condition data of standard colloidal materials (such as epoxy resin) and extracts general features (such as viscosity-temperature relationship and rheological property patterns). When simulating new colloidal materials (such as silicone) or workpiece characteristics (such as curved substrates), the transfer learning module maps the low-dimensional features of the pre-trained model to the new working condition domain through a feature alignment layer, and only fine-tunes the parameters of the fully connected layer to adapt to the new characteristics, avoiding the risk of overfitting caused by training from scratch. AI self-optimization algorithms simulate the rheological behavior of novel colloidal materials through generative adversarial networks (GANs). In a virtual environment, they dynamically adjust the parameters of the pump-valve collaborative control strategy (such as pump displacement slope and valve opening response time). For example, for high-viscosity silicone materials, the algorithm automatically increases the pump displacement slope to compensate for the flow rate reduction caused by the increase in viscosity, while dynamically adjusting the valve opening to reduce pressure fluctuations, thereby achieving adaptive optimization of the control strategy. During parameter adjustment, real-time comparison data of control command execution results and digital twin prediction results are collected. The model's generalization ability is quantified by calculating the prediction error distribution (such as root mean square error and maximum deviation) under new operating conditions. At the same time, the control accuracy (such as the standard deviation of adhesive distribution and trajectory tracking error) before and after optimization is compared, and the control accuracy improvement rate (such as from ±0.2μm to ±0.1μm) is calculated. Finally, a fifth set of physical simulation data is generated, which includes the model's generalization ability (such as prediction accuracy ≥95% across material types) and the control accuracy improvement rate. The fifth physical simulation data is fed back to the hydraulic drive module, and the pump-valve coordinated control strategy is optimized through reinforcement learning algorithm (such as dynamically adjusting PID parameters to adapt to materials with different viscosities). At the same time, it is input into the simulation comprehensive management module for full-domain performance evaluation to verify the adaptive capability of the intelligent decision-making system under multiple working conditions. Feature alignment is used to map the feature space between the source domain (standard materials) and the target domain (new materials). Domain adaptation techniques (such as minimizing the maximum mean difference) are used to reduce the distribution differences between domains and improve the generalization ability of the model under new working conditions. The self-optimizing algorithm continuously collects new working condition data through an online learning mechanism and dynamically updates parameters by combining the transfer learning module. For example, when a change in workpiece characteristics is detected (such as an increase in substrate curvature), the algorithm automatically triggers the transfer learning process, uses the knowledge of the pre-trained model to quickly adapt to the new characteristics, and continuously optimizes the control strategy through a closed-loop feedback mechanism.

[0012] Furthermore, the comprehensive monitoring and diagnostic module includes: A fiber optic grating-quantum tunneling composite sensor array is deployed in hydraulic lines, glue injection head brackets and key mechanical joints. The fiber optic grating achieves strain-temperature dual parameter measurement through the Bragg reflection principle, while the quantum tunneling sensor captures micron-level displacement changes using the quantum tunneling effect. The two work together to achieve submicron-level fault feature extraction. Typical fault modes (such as 0.01mm micro-leakage in hydraulic pipelines and ±0.5% drift of sensors) are actively injected into the virtual testing environment. Multi-physical quantity signals (such as pressure fluctuations, temperature gradients, and vibration spectra) are collected in real time through a composite sensor network. For example, fiber optic gratings monitor the strain changes in hydraulic pipelines, and quantum tunneling sensors capture the minute displacements of the dispensing head bracket, forming a full-dimensional fault feature dataset. Based on the collected sensor data, a fault identification model is constructed using machine learning algorithms (such as support vector machine + random forest). During the training phase, the model learns the feature differences between normal and fault conditions. During the testing phase, simulated faults are classified and verified. The fault identification accuracy is calculated using a confusion matrix (such as true positive rate and false negative rate), and is finally quantified as "fault identification accuracy greater than 98%". After simulating maintenance operations (such as signal calibration and structural reinforcement), the system performance indicators (such as pressure fluctuation suppression rate and trajectory error reduction rate) are compared before and after maintenance. For example, after maintenance, the hydraulic pipeline pressure fluctuation is reduced from ±5% to ±1%. The maintenance efficiency improvement rate is calculated by "(error before maintenance - error after maintenance) / error before maintenance × 100%", and finally quantified as "maintenance efficiency improvement rate greater than 40%". The generated sixth physical simulation data (fault identification accuracy and maintenance efficiency improvement rate) is input into the digital twin module to optimize the parameter thresholds of the fault prediction model (such as adjusting the leakage detection sensitivity) and to provide feedback to adjust the control strategy (such as increasing the trigger threshold of redundant sensors).

[0013] The actual maintenance operation process (such as sensor recalibration and mechanical structure reinforcement) is reproduced in a virtual environment. The system response characteristics after maintenance are monitored through a composite sensor network to verify the effectiveness of the maintenance strategy. For example, the change in the vibration spectrum characteristics of the glue injection head bracket after structural reinforcement verifies the effectiveness of the reinforcement measures in suppressing mechanical vibration. Based on maintenance efficiency improvement data, the maintenance strategy library is iteratively optimized (such as adding adaptive calibration algorithms and optimizing the topology of reinforced structures), and the robustness of the optimized strategies under various fault scenarios is verified in real time through a digital twin module to ensure the accuracy and efficiency of maintenance decisions.

[0014] Furthermore, the global monitoring and diagnostic module also includes: A high-precision sensor array is deployed at the dispensing head, hydraulic lines, and environmental monitoring points. This array includes a fiber optic temperature sensor (resolution ±0.1℃), a quantum tunneling viscosity sensor (measurement range 0.1-1000Pa·s), and a piezoelectric pressure sensor (accuracy ±0.5%FS). By simultaneously collecting data on temperature, viscosity, and pressure, a multi-physical-quantity raw dataset is constructed, providing a foundation for model training. Based on fluid mechanics and thermodynamics, a dynamic coupling equation set of temperature-viscosity-pressure is established. For example, the Arrhenius equation is used to describe the exponential effect of temperature on viscosity, and the Barus equation is used to quantify the thickening effect of pressure on viscosity, forming a nonlinear coupling model of three physical quantities. At the same time, machine learning algorithms (such as random forest regression) are introduced to calibrate the parameters of the coupling equations, thereby improving the prediction accuracy of the model under complex working conditions. During model operation, sensor data is received in real time and the states of the three physical quantities are dynamically updated. By injecting simulated environmental disturbances (such as temperature fluctuations of ±10℃ and vibration acceleration of ±5g), the impact of disturbances on sensor accuracy is analyzed. For example, increased temperature leads to decreased viscosity, which in turn causes fluctuations in the flow rate of the hydraulic system. The model predicts the impact of this process on the injection accuracy through coupled equations and triggers a wide-temperature-range adaptive compensation adjustment control strategy (such as dynamically correcting the pump displacement to offset viscosity changes). In a wide temperature range of -40℃ to 80℃ and under complex vibration scenarios, the model verifies the compensation effect of wide temperature range adaptive compensation. By comparing the monitoring accuracy (e.g., temperature measurement error reduced from ±2℃ to ±0.5℃) and anti-interference ability (e.g., vibration noise suppression rate ≥30dB) before and after compensation, the environmental robustness improvement effect is quantified. Based on the comparison between the model's predicted values ​​and the measured values ​​of reference instruments such as laser interferometers and high-precision viscometers, the monitoring accuracy indicators are calculated. For example, temperature monitoring accuracy is quantified by root mean square error (RMSE) (e.g., ≤0.3℃), viscosity monitoring accuracy is evaluated by relative error, and pressure monitoring accuracy is measured by standard deviation (e.g., ≤0.1MPa). By simulating environmental disturbances of different intensities (such as temperature shock and mechanical vibration), the stability of the model under disturbances was tested. For example, under temperature fluctuations of ±10℃, the error of the glue injection trajectory predicted by the model was still controlled within ±0.1μm. The anti-interference ability was quantified by the disturbance intensity-error increment curve (e.g., error increment ≤5% when temperature fluctuation is ±5℃). The generated seventh physical simulation data (monitoring accuracy, anti-interference capability) is input into the simulation integrated management module to evaluate the stability of the system in a wide temperature range and complex vibration scenarios. For example, through thermodynamic-mechanical coupling analysis, it can guide the optimization of the hydraulic pipeline insulation layer design to reduce the impact of temperature fluctuations, or adjust the sensor layout to improve the anti-vibration interference capability.

[0015] A semi-physical simulation platform control system for an electromechanical-hydraulic coupling dispensing machine includes: Environment Construction Layer: The platform first constructs a virtual test environment based on the quantum-enhanced piezoelectric metamaterial dispensing head and topology optimization model. The dynamic deformation process of the nozzle is simulated through the infrastructure simulation unit of the mechanical execution module, generating the first physical simulation data including the glue distribution and response delay. The physical model parameters, including the nozzle stiffness and driving voltage-displacement response curve in the digital twin, are calibrated. Collaborative Simulation Layer: In the mechanical execution module, a rigid-flexible coupled motion platform and a laser interferometer are used to simulate the high-speed curved trajectory execution process of the dispensing head. Through semi-physical simulation, features such as motion platform pose deviation and mechanical backlash accumulation are captured to generate second physical simulation data, which is transmitted in real time to the global monitoring module as fault prediction input. The physical hardware and digital twin model are run simultaneously to verify the execution effect of virtual debugging commands, and feedback is used to correct the motion platform dynamics model to optimize trajectory planning accuracy. Operating condition verification layer: By combining a variable flow hydraulic pump with a quantum piezoelectric digital pulsation attenuator, the fault state and maintenance operation of the hydraulic system are simulated in the digital hydraulic drive module, generating third physical simulation data including pressure fluctuation suppression effect and energy efficiency recovery rate; a three-dimensional model of mechanical vibration-hydraulic fluctuation-heat conduction is constructed to analyze the impact of mechanical vibration caused by hydraulic shock on glue injection accuracy, and the coupling effect results are generated and input into the digital twin module to verify the robustness of the control algorithm and provide feedback to optimize the glue injection trajectory planning; Real-time optimization layer: A digital twin is built based on the FPGA-quantum computing collaborative architecture to compress the physical model to the scale of real-time computing and generate a fourth physical simulation data containing control command delay and simulation error to support HIL testing; at the same time, an AI self-optimization algorithm enhanced by transfer learning is deployed to simulate the adjustment of characteristic parameters of new colloidal materials or workpieces, generating a fifth physical simulation data with improved model generalization ability and control accuracy, which is fed back to the hydraulic drive module to optimize the pump-valve collaborative control strategy; Diagnostic Assessment Layer: This layer simulates fault states such as hydraulic pipeline micro-leakage and sensor drift, as well as maintenance operations such as signal calibration and structural reinforcement, using a fiber optic grating-quantum tunneling composite sensor network. This generates a sixth physical simulation data set, including fault identification accuracy and maintenance efficiency improvement, to optimize the fault prediction model. A temperature-viscosity-pressure multi-physical quantity fusion monitoring model is constructed to simulate the impact of environmental interference on sensor accuracy, generating a seventh physical simulation data set on monitoring accuracy and anti-interference capability. This data is then input into the simulation integrated management module to evaluate the system's stability in a wide temperature range and complex vibration scenarios. Decision output layer: Integrating the physical simulation data from the first to the seventh stage, the layer performs multi-module data fusion analysis through an edge-cloud collaborative architecture, and outputs simulation evaluation results of the glue volume control accuracy, energy consumption utilization rate, and wide temperature range stability of the equipment under test. The evaluation results are fed back to each module for closed-loop iteration, and finally output optimized glue injection process parameters, control strategies, and maintenance decision suggestions.

[0016] The present invention has the following beneficial effects: In this invention, a closed-loop mechanism with multi-module collaboration is used to solve the mismatch problem. The mechanical execution module generates the first and second physical data on the distribution of adhesive content and pose deviation, calibrates the twin model, and transmits it to the monitoring module. The digital hydraulic drive module generates the third physical data containing pressure fluctuation suppression and drives the mechanical execution. The edge-cloud collaborative digital twin module generates the fourth and fifth physical data containing instruction latency and model accuracy, optimizes the hydraulic strategy, and issues instructions. The global monitoring and diagnosis module generates the sixth and seventh physical data containing fault identification and wide temperature range stability. The simulation comprehensive management module generates optimization suggestions based on the evaluation results and feeds them back to the execution and drive module, effectively solving the problem of dynamic characteristic mismatch between traditional simulation and actual hardware. Attached Figure Description

[0017] Figure 1 This is a system block diagram of a semi-physical simulation platform and control system for an electromechanical-hydraulic coupling dispensing machine proposed in this invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Please see Figure 1 As shown, the present invention is a semi-physical simulation platform and control system for an electromechanical-hydraulic coupling dispensing machine, comprising: A semi-physical simulation platform and control system for an electromechanical-hydraulic coupling dispensing machine, comprising: Mechanical execution module: Constructs a virtual environment for the quantum dispensing head, simulates nozzle deformation, generates first physical data on glue content distribution and response delay, inputs it into the digital twin module to calibrate physical model parameters, indirectly improving the simulation fidelity of the fourth data; integrates a rigid-flexible coupling platform and a laser interferometer to capture high-speed trajectory pose deviation and mechanical clearance characteristics, generates second physical data, and transmits it in real time to the global monitoring and diagnosis module to adjust the model's feature weights and thresholds, optimizing the sixth physical data; Digital hydraulic drive module: Simulates fault and maintenance dynamic response through converter pump and quantum attenuator, generates third physical data including pressure fluctuation suppression and energy efficiency recovery rate, inputs to the global monitoring module to adjust the generation logic of sixth physical data, and synchronizes the parameters to the digital twin module to drive mechanical execution; Edge-cloud collaborative digital twin module: Based on FPGA-quantum architecture, a twin is built to generate fourth physical data containing instruction delay and simulation error. The accuracy of the fault prediction model of the sixth physical data is adjusted. Transfer learning AI algorithm is deployed to simulate the adjustment of new material parameters. Fifth physical data containing model generalization and accuracy improvement rate is generated. The hydraulic control parameter generation logic of the third physical data is adjusted. Instructions are sent to the mechanical execution module. The full-domain monitoring and diagnosis module constructs a multi-physical quantity model through fiber optic-quantum sensors, simulates environmental interference, generates the seventh physical data to evaluate the stability over a wide temperature range, simulates faults and maintenance operations, generates the sixth physical data containing fault identification and maintenance efficiency improvement rates, and adjusts the AI ​​algorithm feature weights of the fifth physical data. Simulation Integrated Management Module: Evaluates the stability of the platform under wide temperature range and complex vibration scenarios. Based on the stability evaluation results, it generates engineering parameter optimization suggestions and feeds them back to the mechanical actuation and digital hydraulic drive modules.

[0020] In one embodiment, the mechanical actuation module includes: Infrastructure Simulation Unit: A virtual testing environment is constructed based on a quantum-enhanced piezoelectric metamaterial dispensing head. A topology optimization model is used to accurately simulate the dynamic deformation process of the nozzle at the sub-micron scale. First, utilizing the piezoelectric effect and quantum tunneling characteristics of the quantum-enhanced piezoelectric metamaterial, a high-precision constitutive model of the nozzle material is established. The design domain is defined as the key deformation regions of the nozzle (such as the outlet section and support structure). Constraints include material stiffness threshold, maximum allowable displacement deviation, and the required glue distribution range. Subsequently, a density-based topology optimization algorithm is used to iteratively optimize the material distribution and structural topology to minimize dynamic deformation error, generating a nozzle geometric model that meets sub-micron precision requirements. This is then simulated using a finite element method (FEM) to simulate the final model. Dispersive element coupling simulation simulates the dynamic response of a nozzle under high-speed motion when subjected to the coupling effects of multiple physical fields such as hydraulic shock and mechanical vibration. It calculates the displacement field, strain field, and colloid flow characteristics at various points on the nozzle surface in real time. Finally, it extracts the colloid volume distribution (spatial distribution of colloid volume flow rate based on fluid dynamics simulation) and response delay (time difference from the issuance of control command to the actual deformation of nozzle) data to form the first physical simulation data containing colloid volume distribution and response delay. This data is directly input into the edge-cloud collaborative digital twin module. The nonlinear characteristic parameters of the nozzle stiffness matrix and driving voltage-displacement response curve in the digital twin are calibrated through parameter mapping algorithm to ensure the consistency between the virtual environment and the actual hardware in submicron-level dynamic characteristics. The physical equipment simulation unit integrates a rigid-flexible coupling motion platform and a laser interferometer to simulate the mechanical execution process of a dispensing head on a high-speed curved trajectory and capture its physical operation characteristics. First, the rigid-flexible coupling motion platform achieves macroscopic trajectory motion through a rigid body module, while the flexible module simulates the minute elastic deformation of the dispensing head support. The two work together to reproduce the motion characteristics under actual working conditions. The laser interferometer measures the motion platform's pose (X / Y / Z axis displacement and rotation angle) and the spatial trajectory of the dispensing head tip in real time, simultaneously acquiring mechanical clearance data (such as joint bearing wear and transmission chain elastic deformation). The measurement signals are processed to calculate the motion platform's pose deviation (root mean square error between the actual and ideal trajectories) and the accumulated mechanical clearance (the amount of mechanical clearance). The statistical characteristics of clearance changes during motion are analyzed, and the force transmission efficiency (the ratio of output force to input driving force, reflecting energy transmission loss) is calculated by combining force sensor data. Finally, a second physical simulation data containing trajectory error and force transmission efficiency is generated. This data is transmitted in real time to the multi-physical quantity fusion global monitoring module via the TSN network as a feature input for the fault prediction model (such as triggering an abnormal mechanical clearance warning when the trajectory error exceeds the threshold). At the same time, it is fed back to the digital twin module, which updates the mass-damping-stiffness parameters of the motion platform through the dynamic model correction algorithm, and optimizes the trajectory planning accuracy of the control commands (such as using the B-spline curve smoothing algorithm to reduce trajectory jitter), forming a closed-loop control link of simulation-measurement-feedback-optimization.

[0021] In one embodiment, the digital hydraulic drive module includes: Hydraulic Fault Dynamic Response Simulation Unit: Through the synergistic effect of a variable flow hydraulic pump and a quantum piezoelectric digital pulsation attenuator, it accurately simulates the fault state and maintenance operations of a hydraulic system. Typical fault modes (such as pump efficiency dropping to below 85% and valve leakage exceeding 0.1 mL / s) are actively injected into the virtual test environment. The pressure fluctuation signal of the hydraulic pipeline is collected in real time through the quantum piezoelectric digital pulsation attenuator. Combined with the flow-pressure characteristic curve of the variable flow pump, the dynamic response characteristics of the hydraulic system under fault conditions are quantified. Maintenance operations such as resetting simulated parameters (e.g., adjusting the gain of the PID controller) and structural adjustments (e.g., optimizing the valve opening) are carried out by adjusting the output flow and pressure of the hydraulic pump in real time through a closed-loop control algorithm and suppressing pressure fluctuations by utilizing the active damping control function of the pulsation attenuator. Based on real-time data collected by the pressure sensor array, the pressure fluctuation suppression effect (such as peak pressure reduction rate and fluctuation frequency attenuation rate) and energy efficiency recovery rate (the ratio of total system energy consumption after maintenance to energy consumption under fault conditions) are calculated, forming a third physical simulation data that includes the pressure fluctuation suppression effect and energy efficiency recovery rate. This data is input into the full-domain monitoring and diagnosis module to verify the effectiveness of the maintenance strategy and is fed back to the digital twin module to adjust the hydraulic impedance matching algorithm (such as optimizing the matching relationship between pipeline diameter, valve opening and pump displacement). Hydraulic-Mechanical-Thermal Dynamic Coupled Simulation Unit: The finite element method is used to establish a mechanical vibration sub-model (simulating the elastic deformation of the glue injection head bracket), a computational fluid dynamics simulation hydraulic fluctuation sub-model (analyzing the hydraulic oil flow characteristics), and a heat conduction sub-model (calculating temperature distribution and viscosity changes). The common node matching of the three domain meshes is achieved through mesh coupling technology to ensure the consistency of boundary conditions. During the glue injection process, the mechanical vibrations caused by hydraulic shock (such as the vibration of the glue injection head caused by sudden pressure changes when the pump starts and stops), the thermal effects caused by vibration (such as frictional heat generation changing the viscosity of hydraulic oil), and the influence of heat conduction on mechanical properties (such as the decrease in material stiffness caused by temperature rise) are simulated. The interaction of each physical field is calculated in real time through a multi-physics coupling solver to generate coupling effect analysis results (such as the influence coefficient of vibration frequency on glue injection accuracy and the influence curve of temperature gradient on hydraulic system efficiency). The coupling effect analysis results are input into the digital twin module to verify the robustness of the control algorithm under complex working conditions (such as anti-interference ability and trajectory tracking accuracy), and fed back to the mechanical execution module to optimize the glue injection trajectory planning (such as avoiding vibration-sensitive areas and adjusting the glue injection speed to reduce the impact of thermal effects).

[0022] In one embodiment, the edge-cloud collaborative digital twin electronic control module includes: A digital twin is constructed based on an FPGA-quantum computing collaborative architecture. The FPGA, as a classical computing acceleration unit, is responsible for real-time processing of sensor data, execution of control algorithms and communication protocol stacks. It simulates complex nonlinear physical processes (such as high-order elastic deformation of nozzle materials and quantum tunneling effect of hydraulic systems) through qubits. The two achieve data interaction and task splitting through a high-speed bus. For example, the FPGA processes TSN network data packets synchronized at the sub-microsecond level, while the quantum computing module optimizes the nonlinear parameter calibration of the nozzle stiffness matrix in the digital twin. First, based on the physical properties of the quantum-enhanced piezoelectric metamaterial dispensing head, a quantum-state-described nozzle dynamics model is established in the quantum computing module. Subsequently, the FPGA is used to deploy the model in real time through a hardware description language, forming a hybrid computing architecture of quantum modeling and classical acceleration.

[0023] For high-dimensional physical models (such as finite element models with millions of degrees of freedom), a model reduction method based on POD (orthogonal eigenvalue decomposition) is adopted to extract the minority modes that dominate the dynamic characteristics (such as the first 10 vibration modes), thus compressing the model size to a scale that can be computed in real time (such as from 10). 6The degree of freedom is reduced to 10³ degrees of freedom. At the same time, the nonlinear part is approximated by combining the Krylov subspace method. The reduced-order model is deployed through FPGA hardware accelerator and uses parallel computing units to achieve more than one million model updates per second, supporting the real-time simulation requirements in HIL testing.

[0024] During the operation of the digital twin, the FPGA collects the timestamps of the control command issuance and the actual hardware execution in real time, and calculates the control command delay (such as the time difference from the FPGA issuing the command to the dispensing head starting to move); at the same time, by comparing the nozzle displacement predicted by the digital twin with the displacement data measured by the laser interferometer, the simulation error is quantified (such as root mean square error ≤ 0.05μm). The generated fourth physical simulation data includes indicators such as control command delay and simulation error, which directly support the verification and virtual debugging of control strategies in HIL testing. For example, if the simulation error exceeds the threshold, the model correction process is triggered, and the nozzle stiffness parameters are recalibrated through the quantum computing module. At the same time, the data is input into the global monitoring module to evaluate the real-time performance of the control strategy (such as whether the command delay meets the ±1μs process requirement) and fed back to the mechanical execution module to optimize the dispensing head drive voltage curve (such as adjusting the PWM waveform parameters to reduce response delay).

[0025] In one embodiment, the edge-cloud collaborative digital twin electronic control module further includes: Based on pre-trained deep learning models (such as convolutional neural networks + long short-term memory networks), an integrated transfer learning module is used to transfer knowledge across working conditions. The pre-trained model completes initial training on working condition data of standard colloidal materials (such as epoxy resin) and extracts general features (such as viscosity-temperature relationship and rheological property patterns). When simulating new colloidal materials (such as silicone) or workpiece characteristics (such as curved substrates), the transfer learning module maps the low-dimensional features of the pre-trained model to the new working condition domain through a feature alignment layer, and only fine-tunes the parameters of the fully connected layer to adapt to the new characteristics, avoiding the risk of overfitting caused by training from scratch. AI self-optimization algorithms simulate the rheological behavior of novel colloidal materials through generative adversarial networks (GANs). In a virtual environment, they dynamically adjust the parameters of the pump-valve collaborative control strategy (such as pump displacement slope and valve opening response time). For example, for high-viscosity silicone materials, the algorithm automatically increases the pump displacement slope to compensate for the flow rate reduction caused by the increase in viscosity, while dynamically adjusting the valve opening to reduce pressure fluctuations, thereby achieving adaptive optimization of the control strategy. During parameter adjustment, real-time comparison data of control command execution results and digital twin prediction results are collected. The model's generalization ability is quantified by calculating the prediction error distribution (such as root mean square error and maximum deviation) under new operating conditions. At the same time, the control accuracy (such as the standard deviation of adhesive distribution and trajectory tracking error) before and after optimization is compared, and the control accuracy improvement rate (such as from ±0.2μm to ±0.1μm) is calculated. Finally, a fifth set of physical simulation data is generated, which includes the model's generalization ability (such as prediction accuracy ≥95% across material types) and the control accuracy improvement rate. The fifth physical simulation data is fed back to the hydraulic drive module, and the pump-valve coordinated control strategy is optimized through reinforcement learning algorithm (such as dynamically adjusting PID parameters to adapt to materials with different viscosities). At the same time, it is input into the simulation comprehensive management module for full-domain performance evaluation to verify the adaptive capability of the intelligent decision-making system under multiple working conditions. Feature alignment is used to map the feature space between the source domain (standard materials) and the target domain (new materials). Domain adaptation techniques (such as minimizing the maximum mean difference) are used to reduce the distribution differences between domains and improve the generalization ability of the model under new working conditions. The self-optimizing algorithm continuously collects new working condition data through an online learning mechanism and dynamically updates parameters by combining the transfer learning module. For example, when a change in workpiece characteristics is detected (such as an increase in substrate curvature), the algorithm automatically triggers the transfer learning process, uses the knowledge of the pre-trained model to quickly adapt to the new characteristics, and continuously optimizes the control strategy through a closed-loop feedback mechanism.

[0026] In one embodiment, the global monitoring and diagnostic module includes: A fiber optic grating-quantum tunneling composite sensor array is deployed in hydraulic lines, glue injection head brackets and key mechanical joints. The fiber optic grating achieves strain-temperature dual parameter measurement through the Bragg reflection principle, while the quantum tunneling sensor captures micron-level displacement changes using the quantum tunneling effect. The two work together to achieve submicron-level fault feature extraction. Typical fault modes (such as 0.01mm micro-leakage in hydraulic pipelines and ±0.5% drift of sensors) are actively injected into the virtual testing environment. Multi-physical quantity signals (such as pressure fluctuations, temperature gradients, and vibration spectra) are collected in real time through a composite sensor network. For example, fiber optic gratings monitor the strain changes in hydraulic pipelines, and quantum tunneling sensors capture the minute displacements of the dispensing head bracket, forming a full-dimensional fault feature dataset. Based on the collected sensor data, a fault identification model is constructed using machine learning algorithms (such as support vector machine + random forest). During the training phase, the model learns the feature differences between normal and fault conditions. During the testing phase, simulated faults are classified and verified. The fault identification accuracy is calculated using a confusion matrix (such as true positive rate and false negative rate), and is finally quantified as "fault identification accuracy greater than 98%". After simulating maintenance operations (such as signal calibration and structural reinforcement), the system performance indicators (such as pressure fluctuation suppression rate and trajectory error reduction rate) are compared before and after maintenance. For example, after maintenance, the hydraulic pipeline pressure fluctuation is reduced from ±5% to ±1%. The maintenance efficiency improvement rate is calculated by "(error before maintenance - error after maintenance) / error before maintenance × 100%", and finally quantified as "maintenance efficiency improvement rate greater than 40%". The generated sixth physical simulation data (fault identification accuracy and maintenance efficiency improvement rate) is input into the digital twin module to optimize the parameter thresholds of the fault prediction model (such as adjusting the leakage detection sensitivity) and to provide feedback to adjust the control strategy (such as increasing the trigger threshold of redundant sensors).

[0027] The actual maintenance operation process (such as sensor recalibration and mechanical structure reinforcement) is reproduced in a virtual environment. The system response characteristics after maintenance are monitored through a composite sensor network to verify the effectiveness of the maintenance strategy. For example, the change in the vibration spectrum characteristics of the glue injection head bracket after structural reinforcement verifies the effectiveness of the reinforcement measures in suppressing mechanical vibration. Based on maintenance efficiency improvement data, the maintenance strategy library is iteratively optimized (such as adding adaptive calibration algorithms and optimizing the topology of reinforced structures), and the robustness of the optimized strategies under various fault scenarios is verified in real time through a digital twin module to ensure the accuracy and efficiency of maintenance decisions.

[0028] In one embodiment, the global monitoring and diagnostic module further includes: A high-precision sensor array is deployed at the dispensing head, hydraulic lines, and environmental monitoring points. This array includes a fiber optic temperature sensor (resolution ±0.1℃), a quantum tunneling viscosity sensor (measurement range 0.1-1000Pa·s), and a piezoelectric pressure sensor (accuracy ±0.5%FS). By simultaneously collecting data on temperature, viscosity, and pressure, a multi-physical-quantity raw dataset is constructed, providing a foundation for model training. Based on fluid mechanics and thermodynamics, a dynamic coupling equation set of temperature-viscosity-pressure is established. For example, the Arrhenius equation is used to describe the exponential effect of temperature on viscosity, and the Barus equation is used to quantify the thickening effect of pressure on viscosity, forming a nonlinear coupling model of three physical quantities. At the same time, machine learning algorithms (such as random forest regression) are introduced to calibrate the parameters of the coupling equations, thereby improving the prediction accuracy of the model under complex working conditions. During model operation, sensor data is received in real time and the states of the three physical quantities are dynamically updated. By injecting simulated environmental disturbances (such as temperature fluctuations of ±10℃ and vibration acceleration of ±5g), the impact of disturbances on sensor accuracy is analyzed. For example, increased temperature leads to decreased viscosity, which in turn causes fluctuations in the flow rate of the hydraulic system. The model predicts the impact of this process on the injection accuracy through coupled equations and triggers a wide-temperature-range adaptive compensation adjustment control strategy (such as dynamically correcting the pump displacement to offset viscosity changes). In a wide temperature range of -40℃ to 80℃ and under complex vibration scenarios, the model verifies the compensation effect of wide temperature range adaptive compensation. By comparing the monitoring accuracy (e.g., temperature measurement error reduced from ±2℃ to ±0.5℃) and anti-interference ability (e.g., vibration noise suppression rate ≥30dB) before and after compensation, the environmental robustness improvement effect is quantified. Based on the comparison between the model's predicted values ​​and the measured values ​​of reference instruments such as laser interferometers and high-precision viscometers, the monitoring accuracy indicators are calculated. For example, temperature monitoring accuracy is quantified by root mean square error (RMSE) (e.g., ≤0.3℃), viscosity monitoring accuracy is evaluated by relative error, and pressure monitoring accuracy is measured by standard deviation (e.g., ≤0.1MPa). By simulating environmental disturbances of different intensities (such as temperature shock and mechanical vibration), the stability of the model under disturbances was tested. For example, under temperature fluctuations of ±10℃, the error of the glue injection trajectory predicted by the model was still controlled within ±0.1μm. The anti-interference ability was quantified by the disturbance intensity-error increment curve (e.g., error increment ≤5% when temperature fluctuation is ±5℃). The generated seventh physical simulation data (monitoring accuracy, anti-interference capability) is input into the simulation integrated management module to evaluate the stability of the system in a wide temperature range and complex vibration scenarios. For example, through thermodynamic-mechanical coupling analysis, it can guide the optimization of the hydraulic pipeline insulation layer design to reduce the impact of temperature fluctuations, or adjust the sensor layout to improve the anti-vibration interference capability.

[0029] A semi-physical simulation platform control system for an electromechanical-hydraulic coupling dispensing machine includes: Environment Construction Layer: The platform first constructs a virtual test environment based on the quantum-enhanced piezoelectric metamaterial dispensing head and topology optimization model. The dynamic deformation process of the nozzle is simulated through the infrastructure simulation unit of the mechanical execution module, generating the first physical simulation data including the glue distribution and response delay. The physical model parameters, including the nozzle stiffness and driving voltage-displacement response curve in the digital twin, are calibrated. Collaborative Simulation Layer: In the mechanical execution module, a rigid-flexible coupled motion platform and a laser interferometer are used to simulate the high-speed curved trajectory execution process of the dispensing head. Through semi-physical simulation, features such as motion platform pose deviation and mechanical backlash accumulation are captured to generate second physical simulation data, which is transmitted in real time to the global monitoring module as fault prediction input. The physical hardware and digital twin model are run simultaneously to verify the execution effect of virtual debugging commands, and feedback is used to correct the motion platform dynamics model to optimize trajectory planning accuracy. Operating condition verification layer: By combining a variable flow hydraulic pump with a quantum piezoelectric digital pulsation attenuator, the fault state and maintenance operation of the hydraulic system are simulated in the digital hydraulic drive module, generating third physical simulation data including pressure fluctuation suppression effect and energy efficiency recovery rate; a three-dimensional model of mechanical vibration-hydraulic fluctuation-heat conduction is constructed to analyze the impact of mechanical vibration caused by hydraulic shock on glue injection accuracy, and the coupling effect results are generated and input into the digital twin module to verify the robustness of the control algorithm and provide feedback to optimize the glue injection trajectory planning; Real-time optimization layer: A digital twin is built based on the FPGA-quantum computing collaborative architecture to compress the physical model to the scale of real-time computing and generate a fourth physical simulation data containing control command delay and simulation error to support HIL testing; at the same time, an AI self-optimization algorithm enhanced by transfer learning is deployed to simulate the adjustment of characteristic parameters of new colloidal materials or workpieces, generating a fifth physical simulation data with improved model generalization ability and control accuracy, which is fed back to the hydraulic drive module to optimize the pump-valve collaborative control strategy; Diagnostic Assessment Layer: This layer simulates fault states such as hydraulic pipeline micro-leakage and sensor drift, as well as maintenance operations such as signal calibration and structural reinforcement, using a fiber optic grating-quantum tunneling composite sensor network. This generates a sixth physical simulation data set, including fault identification accuracy and maintenance efficiency improvement, to optimize the fault prediction model. A temperature-viscosity-pressure multi-physical quantity fusion monitoring model is constructed to simulate the impact of environmental interference on sensor accuracy, generating a seventh physical simulation data set on monitoring accuracy and anti-interference capability. This data is then input into the simulation integrated management module to evaluate the system's stability in a wide temperature range and complex vibration scenarios. Decision output layer: Integrating the physical simulation data from the first to the seventh stage, the layer performs multi-module data fusion analysis through an edge-cloud collaborative architecture, and outputs simulation evaluation results of the glue volume control accuracy, energy consumption utilization rate, and wide temperature range stability of the equipment under test. The evaluation results are fed back to each module for closed-loop iteration, and finally output optimized glue injection process parameters, control strategies, and maintenance decision suggestions.

[0030] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A semi-physical simulation platform for electromechanical-hydraulic coupling of a dispensing machine, characterized in that, include: Mechanical execution module: Constructs a virtual environment for the quantum dispensing head, simulates nozzle deformation, generates first physical data on glue content distribution and response delay, inputs it into the digital twin module to calibrate physical model parameters, indirectly improving the simulation fidelity of the fourth data; integrates a rigid-flexible coupling platform and a laser interferometer to capture high-speed trajectory pose deviation and mechanical clearance characteristics, generates second physical data, and transmits it in real time to the global monitoring and diagnosis module to adjust the model's feature weights and thresholds, optimizing the sixth physical data; Digital hydraulic drive module: Simulates fault and maintenance dynamic response through converter pump and quantum attenuator, generates third physical data including pressure fluctuation suppression and energy efficiency recovery rate, inputs to the global monitoring module to adjust the generation logic of sixth physical data, and synchronizes the parameters to the digital twin module to drive mechanical execution; Edge-cloud collaborative digital twin module: Based on FPGA-quantum architecture, a twin is built to generate fourth physical data containing instruction delay and simulation error. The accuracy of the fault prediction model of the sixth physical data is adjusted. Transfer learning AI algorithm is deployed to simulate the adjustment of new material parameters. Fifth physical data containing model generalization and accuracy improvement rate is generated. The hydraulic control parameter generation logic of the third physical data is adjusted. Instructions are sent to the mechanical execution module. The full-domain monitoring and diagnosis module constructs a multi-physical quantity model through fiber optic-quantum sensors, simulates environmental interference, generates the seventh physical data to evaluate the stability over a wide temperature range, simulates faults and maintenance operations, generates the sixth physical data containing fault identification and maintenance efficiency improvement rates, and adjusts the AI ​​algorithm feature weights of the fifth physical data. Simulation Integrated Management Module: Evaluates the stability of the platform under wide temperature range and complex vibration scenarios. Based on the stability evaluation results, it generates engineering parameter optimization suggestions and feeds them back to the mechanical actuation and digital hydraulic drive modules.

2. The electromechanical-hydraulic coupling semi-physical simulation platform for a dispensing machine according to claim 1, characterized in that, The mechanical actuation module includes: Infrastructure simulation unit: A virtual test environment is constructed based on the quantum-enhanced piezoelectric metamaterial dispensing head. The dynamic deformation process of the nozzle is simulated through a topology optimization model, generating the first physical simulation data containing glue distribution and response delay. This data is then input into the edge-cloud collaborative digital twin module to calibrate the physical model parameters in the digital twin. Physical equipment simulation unit: Integrating a rigid-flexible coupling motion platform and a laser interferometer, it simulates the mechanical execution process of the dispensing head on a high-speed curved trajectory, captures physical operation characteristics including motion platform pose deviation and mechanical backlash accumulation, and outputs second physical simulation data including trajectory error and force transmission efficiency. This data is transmitted in real time to the multi-physical quantity fusion global monitoring module as input features for the fault prediction model.

3. The electro-hydraulic coupling semi-physical simulation platform for dispensing adhesives according to claim 1, characterized in that, The digital hydraulic drive module includes: Hydraulic Fault Dynamic Response Simulation Unit: By combining a variable flow hydraulic pump with a quantum piezoelectric digital pulsation attenuator, the dynamic response of the hydraulic system under fault conditions and maintenance operations is simulated, generating third-party physical simulation data including pressure fluctuation suppression effect and energy efficiency recovery rate. This data is then input into the full-domain monitoring and diagnostic module to verify the effectiveness of maintenance strategies and optimize the fault prediction model. Hydraulic-Mechanical-Thermal Dynamic Coupled Simulation Unit: Constructs a three-dimensional coupled simulation model of mechanical vibration, hydraulic fluctuation and heat conduction to simulate the impact of multi-physics field interaction on the control strategy during glue injection, generates coupling effect analysis results, inputs them to the digital twin module to verify the robustness of the control algorithm under complex working conditions, and feeds them back to the mechanical execution module to optimize glue injection trajectory planning.

4. The electromechanical-hydraulic coupling semi-physical simulation platform for a dispensing machine according to claim 1, characterized in that, The edge-cloud collaborative digital twin electronic control module includes: Edge-Cloud Collaborative Twin Infrastructure Simulation Unit: Based on the FPGA-quantum computing collaborative architecture, a digital twin is constructed. By reducing the model order, the physical model is compressed to a scale that can be computed in real time. Synchronizing with the actual hardware, it generates a fourth physical simulation data that includes control command delays and simulation errors. It supports HIL testing and virtual debugging, and feeds back to the mechanical execution module to optimize the dispensing head drive voltage curve. At the same time, it is input to the global monitoring module to evaluate the real-time performance of the control strategy.

5. The electromechanical-hydraulic coupling semi-physical simulation platform for a dispensing machine according to claim 1, characterized in that, The edge-cloud collaborative digital twin electronic control module also includes: Edge-Cloud Collaborative AI Self-Optimization Physical Simulation Unit: Deploys AI self-optimization algorithms enhanced by transfer learning to simulate the parameter adjustment process under the characteristics of novel colloidal materials or workpieces. It outputs fifth physical simulation data containing model generalization ability and control accuracy improvement rate to verify the adaptive capability of the intelligent decision-making system and feeds it back to the hydraulic drive module to optimize the pump-valve collaborative control strategy.

6. The electromechanical-hydraulic coupling semi-physical simulation platform for a dispensing machine according to claim 1, characterized in that, The comprehensive monitoring and diagnostic module includes: Maintenance Auxiliary Simulation Unit: Simulates fault states and corresponding maintenance operations through a fiber optic grating-quantum tunneling composite sensor network, generates sixth physical simulation data including fault identification accuracy and maintenance efficiency improvement rate, optimizes the fault prediction model, and feeds it back to the digital twin module to adjust the control strategy.

7. The electromechanical-hydraulic coupling semi-physical simulation platform for a dispensing machine according to claim 1, characterized in that, The global monitoring and diagnostic module also includes: Multi-physics coupling unit: Construct a multi-physical quantity fusion monitoring model of temperature, viscosity and pressure, simulate the impact of environmental interference on sensor accuracy, verify the environmental robustness of the wide-temperature-range adaptive compensation unit, generate the seventh physical simulation data including monitoring accuracy and anti-interference capability, and input it into the simulation comprehensive management module to evaluate the stability of the system in a wide temperature range and complex vibration scenarios.

8. The control system used in the electromechanical-hydraulic coupling semi-physical simulation platform for a dispensing machine according to claim 1, characterized in that, include: Environment Construction Layer: The platform first constructs a virtual test environment based on the quantum-enhanced piezoelectric metamaterial dispensing head and topology optimization model. The dynamic deformation process of the nozzle is simulated through the infrastructure simulation unit of the mechanical execution module, generating the first physical simulation data including the glue distribution and response delay. The physical model parameters, including the nozzle stiffness and driving voltage-displacement response curve in the digital twin, are calibrated. Collaborative Simulation Layer: In the mechanical execution module, a rigid-flexible coupled motion platform and a laser interferometer are used to simulate the high-speed curved trajectory execution process of the dispensing head. Through semi-physical simulation, features such as motion platform pose deviation and mechanical backlash accumulation are captured to generate second physical simulation data, which is transmitted in real time to the global monitoring module as fault prediction input. The physical hardware and digital twin model are run simultaneously to verify the execution effect of virtual debugging commands, and feedback is used to correct the motion platform dynamics model to optimize trajectory planning accuracy. Operating condition verification layer: By combining a variable flow hydraulic pump with a quantum piezoelectric digital pulsation attenuator, the fault state and maintenance operation of the hydraulic system are simulated in the digital hydraulic drive module, generating third physical simulation data including pressure fluctuation suppression effect and energy efficiency recovery rate; a three-dimensional model of mechanical vibration-hydraulic fluctuation-heat conduction is constructed to analyze the impact of mechanical vibration caused by hydraulic shock on glue injection accuracy, and the coupling effect results are generated and input into the digital twin module to verify the robustness of the control algorithm and provide feedback to optimize the glue injection trajectory planning; Real-time optimization layer: A digital twin is built based on the FPGA-quantum computing collaborative architecture to compress the physical model to the scale of real-time computing and generate a fourth physical simulation data containing control command delay and simulation error to support HIL testing; at the same time, an AI self-optimization algorithm enhanced by transfer learning is deployed to simulate the adjustment of characteristic parameters of new colloidal materials or workpieces, generating a fifth physical simulation data with improved model generalization ability and control accuracy, which is fed back to the hydraulic drive module to optimize the pump-valve collaborative control strategy; Diagnostic Assessment Layer: This layer simulates fault states such as hydraulic pipeline micro-leakage and sensor drift, as well as maintenance operations such as signal calibration and structural reinforcement, using a fiber optic grating-quantum tunneling composite sensor network. This generates a sixth physical simulation data set, including fault identification accuracy and maintenance efficiency improvement, to optimize the fault prediction model. A temperature-viscosity-pressure multi-physical quantity fusion monitoring model is constructed to simulate the impact of environmental interference on sensor accuracy, generating a seventh physical simulation data set on monitoring accuracy and anti-interference capability. This data is then input into the simulation integrated management module to evaluate the system's stability in a wide temperature range and complex vibration scenarios. Decision output layer: Integrating the physical simulation data from the first to the seventh stage, the layer performs multi-module data fusion analysis through an edge-cloud collaborative architecture, and outputs simulation evaluation results of the glue volume control accuracy, energy consumption utilization rate, and wide temperature range stability of the equipment under test. The evaluation results are fed back to each module for closed-loop iteration, and finally output optimized glue injection process parameters, control strategies, and maintenance decision suggestions.

Images