Digital twinning-based heat-assisted flow drilling and riveting control method and system

Abstract

The invention discloses a digital twinning-based heat-assisted flow drilling and riveting control method and a digital twinning-based heat-assisted flow drilling and riveting control system. The method comprises the following steps: acquiring temperature, pressure, displacement and surface topography data; a temperature-pressure-morphology comprehensive index is constructed through weighted fusion; real-time processing is carried out through the FPGA chip and the lightweight AI model; a Pareto frontier solution set is generated in combination with a Monte Carlo-NSGA-II hybrid algorithm; constructing a heat-force-phase change coupling unified constitutive model to carry out multi-physics field reverse optimization; dynamic updating of process parameters is realized through a digital twinborn case library; and dynamically modulating PID parameters by adopting deep reinforcement learning. The system comprises a multi-source data acquisition terminal, an intelligent decision module and a dynamic control module. According to the method, the problems of parameter lag and poor adaptability of a traditional method are solved, and the processing task can be efficiently and accurately completed.

Description

Technical Field

[0001] The present invention relates to the field of precision machining control technology, and specifically to a heat-assisted flow drilling and riveting control method and system based on digital twins, which is suitable for the connection processing of metal parts. Background Art

[0002] Traditional heat-assisted flow drilling and riveting face many challenges in the actual production and processing process. First, sensor data are often independent of each other, lacking multi-modal fusion judgment, resulting in the inability to accurately and real-time detect abnormal conditions in the production and processing process. Secondly, empirical parameters are used in the control algorithm, but when faced with complex processing material structures (such as the laminated structure of carbon fiber materials and titanium alloys), its adaptability is significantly reduced. Furthermore, although the cloud-based centralized computing mode provides powerful data processing capabilities, the resulting lag in control instructions cannot be ignored, and the lag caused during high-speed processing reduces processing stability. Finally, the process is highly dependent on trial and error, which can only be achieved at a relatively high cost and a long adjustment cycle, and it is difficult to meet the needs of modern manufacturing for efficient and flexible production. Summary of the invention

[0003] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide a heat-assisted flow drilling and riveting control method and system based on digital twins. The present invention can complete processing tasks efficiently and accurately, and at the same time has good maintainability and scalability, providing a new technical solution for the field of intelligent manufacturing.

[0004] The present invention is implemented by the following technical solution: a heat-assisted flow drilling and riveting control method based on digital twins, comprising the following steps: Step 1: Collect processing data of heat-assisted flow drilling and riveting equipment, including temperature, pressure, displacement and surface morphology image data of the processing area; Step 2: preprocess the temperature, pressure, displacement and surface morphology image data, standardize the preprocessed temperature, pressure and surface morphology image data, and then perform weighted fusion on the standardized temperature, pressure and surface morphology image data to obtain temperature-pressure-morphology weighted fusion data; Step 3: Use FPGA chips and lightweight AI models to process standardized data and temperature-pressure-morphology weighted fusion data in real time, dynamically optimize process parameters, and simultaneously generate temperature prediction and defect detection results; Step 4: Based on the process parameters, temperature prediction and defect detection results obtained in step 3, further optimization is performed through the Monte Carlo-NSGA-II hybrid algorithm to generate candidate process parameters, screen the Pareto frontier solution set, and transmit it to the digital twin case library in real time; Step 5: Construct a unified constitutive model of thermal-mechanical-phase change coupling, and use the multi-physics field coupling numerical analysis method to reversely optimize the Pareto frontier solution set. The Pareto frontier solution set is re-entered into the lightweight AI model through the digital twin case library for re-optimization, and the real-time monitored processing data is reversely input into the simulation model for closed-loop verification; Step 6: The verified process parameters are included in the digital twin case library, and the digital twin case library pushes a new Pareto frontier solution set to the lightweight AI model; Step 7: Dynamically modulate the PID parameters through the deep reinforcement learning model, combine with the verified process parameters, and control the heat-assisted flow drilling and riveting equipment through the PID control circuit.

[0005] Further preferably, the temperature-pressure-morphology weighted fusion data , is the normalized temperature data, F * is the normalized pressure data, I * It is the surface topography image data after binarization.

[0006] Further preferably, in step 7, when the PID parameters are dynamically modulated by the deep reinforcement learning model, the PID parameters are corrected in real time in combination with the tensile strength-energy consumption balance reward function, and hardware-level protection is triggered when the limit is exceeded, and the control results of the heat-assisted flow drilling and riveting equipment are fed back to the digital twin case library.

[0007] Further preferably, the state space of the deep reinforcement learning model (DRL) includes real-time processing data and digital twin input, and the action space is divided into continuous actions and discrete actions.

[0008] Further preferably, the tensile strength-energy consumption balance reward function is: R=10×S 达标 +5×E 降低 −20×T 超限 −15×V 超标 , where R is the reward value, S 达标 Indicates that the connection strength meets the standard, E 降低 Indicates reduced energy consumption, T 超限 Indicates temperature over limit mark; V 超标 Indicates that the vibration amplitude exceeds the limit.

[0009] Further preferably, the process of step 3 is as follows: Step b1: Build a lightweight AI model architecture; Step b2: Split the lightweight AI model layer into parallel pipeline tasks, match the FPGA chip, and use the FPGA chip to cache pre-stored data; Step b3: The loss function is obtained by summing the mean square error and the structural similarity according to the weights; Step b4: dynamically optimizing process parameters; Step b5: Generate temperature prediction and defect detection results.

[0010] Further preferably, according to the normalized temperature data And the normalized pressure data Predicted target temperature T opt ; Use lightweight AI model to binarize surface morphology image data I * Classify defect types, output defect coordinates, types, and severity levels, and provide real-time feedback to the parameter optimizer.

[0011] The present invention also provides a digital twin-based heat-assisted flow drilling and riveting control system, comprising: Multi-source data acquisition terminal, used to capture temperature, pressure, displacement and surface topography image data of the processing area in real time; Intelligent decision-making module, which consists of a lightweight AI model, parameter optimizer, digital twin case library and digital twin model library. The digital twin model library is used to store the thermal-mechanical-phase change coupled unified constitutive model and process data through the Monte Carlo-NSGA-II hybrid algorithm; The dynamic control module is composed of a safety monitoring module, a PID control module, and a servo motor and a heating module of the heat-assisted flow drilling and riveting equipment; the PID control module is composed of a PID control circuit and a deep reinforcement learning model, and the PID parameters are dynamically modulated through the deep reinforcement learning model.

[0012] Further preferably, the multi-source data acquisition terminal includes an infrared thermal imager, a multi-axis force sensor, a laser displacement meter and an industrial camera.

[0013] Compared with the existing control method, the present invention has the following technical effects: Temperature, pressure, displacement and surface morphology data are collected through infrared thermal imagers, multi-axis force sensors and other equipment. After pre-processing such as Gaussian filtering and FFT transformation, weighted fusion is used to achieve comprehensive perception of the machining process and improve the level of precise control.

[0014] Based on FPGA chips and lightweight AI models, combined with neural architecture search and hardware co-optimization, temperature prediction and defect detection with a delay of ≤10ms can be achieved.

[0015] The Monte Carlo-NSGA-II algorithm is used to generate the Pareto front solution set (tensile strength ≥ 800MPa, energy consumption ≤ 500J / time), and the thermal-mechanical-phase change coupled unified constitutive model is combined to perform multi-physics field simulation to reduce the error rate of heat affected zone morphology prediction.

[0016] By storing optimized process parameters and simulation data in the digital twin case library, the new product process debugging cycle is shortened and rapid process reuse and expansion are supported.

[0017] The deep reinforcement learning model dynamically modulates PID parameters and combines the tensile strength-energy consumption balance reward function to achieve high-precision closed-loop control of the machining process. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 Schematic diagram of the heat-assisted flow drilling and riveting control system based on digital twin; Figure 2 This is a flow chart of the thermally assisted flow drilling and riveting control method based on digital twin; Figure 3 This is the circuit design diagram of the PID control module. DETAILED DESCRIPTION

[0019] In order to facilitate those skilled in the art to understand and implement the present invention, the present invention is further described in detail and in depth below with reference to the accompanying drawings.

[0020] like Figure 1 As shown in the figure, the heat-assisted flow drilling and riveting control system based on digital twin includes: The multi-source data acquisition terminal, including infrared thermal imager, multi-axis force sensor, laser displacement meter and industrial camera, is responsible for real-time capture of temperature, pressure, displacement and surface topography image data of the processing area. The configured infrared thermal imager is used for temperature field monitoring, monitoring temperature at a sampling rate of 100Hz, with a spatial resolution of 0.1mm² / pixel, a temperature range of 400-700℃, and an accuracy of ±1℃. The multi-axis force sensor, integrated in the spindle of the drilling and riveting mechanism, is used to monitor pressure and vibration, with a range of ±5kN, a bandwidth of 10kHz, and data transmission via the Ether CAT protocol. The laser displacement meter is used to measure the surface deformation and displacement of the workpiece, using triangulation technology with an accuracy of ±0.02mm. The industrial camera, with a frame rate of 120fps and a resolution of 2048×1536 pixels, is used to capture surface topography image data.

[0021] The intelligent decision-making module consists of a lightweight AI model, a parameter optimizer, a digital twin case library and a digital twin model library; the digital twin model library is used to store numerical models such as the thermal-mechanical-phase change coupled unified constitutive model and the Monte Carlo-NSGA-II hybrid algorithm.

[0022] The dynamic control module consists of a safety monitoring module, a PID control module, and a servo motor and heating module for the heat-assisted flow drilling and riveting equipment. The PID control module dynamically modulates the PID parameters in combination with the deep reinforcement learning model (DRL); detects temperature and pressure over-limit events in real time and triggers the abnormality handling mechanism.

[0023] As attached Figure 2 As shown, the digital twin-based heat-assisted flow drilling and riveting control method proposed in the present invention adopts a modular hierarchical architecture and divides the implementation process into a three-loop closed-loop system of "perception-decision-execution", which is specifically divided into: multimodal perception layer (step 1-step 3), intelligent decision-making layer (step 4-step 6), and dynamic control layer (step 7).

[0024] Step 1: Collect data through a multi-source data acquisition terminal to collect processing data of the heat-assisted flow drilling and riveting equipment, including temperature, pressure, displacement and surface morphology image data of the processing area; Step 2: preprocess the temperature, pressure, displacement and surface morphology image data, standardize the preprocessed temperature, pressure and surface morphology image data, and then perform weighted fusion on the standardized temperature, pressure and surface morphology image data to obtain temperature-pressure-morphology weighted fusion data; Preprocessing refers to preprocessing each set of processing data through noise suppression, feature extraction, smooth trajectory and edge detection to obtain high-quality basic data. Preprocessing can improve the credibility of single sensor data, and through multi-source data fusion, it can achieve comprehensive perception and precise control of the processing process, promote process efficiency improvement (debugging cycle shortened by 40%) and processing quality optimization (defect rate ≤ 1%). The specific preprocessing methods are as follows: Temperature data: Gaussian filtering (window size: 3×3) was used to remove noise and retain effective thermal field characteristics; Pressure data: Fast Fourier transform (FFT) extracts the characteristic spectrum of the 0.1-5kHz frequency band and identifies the vibration characteristic frequency band; Displacement data: Kalman filtering smoothes the trajectory and reduces measurement jitter errors; Surface morphology image data: Canny edge detection is performed through the OpenCV library to mark defect areas.

[0025] Standardization includes: Normalized temperature data: ∈[0,1]; Normalized pressure data: ∈[−1,1]; Binarized surface topography image data: I * ∈{0,1}(the defective area is marked as 1, and the others are marked as 0).

[0026] In the formula, is the normalized temperature data, F * is the normalized pressure data, I * is the surface topography image data after binarization, T max , T min The maximum and minimum temperatures are set respectively.max is the set maximum pressure, T is the temperature data after preprocessing, and F is the pressure data after preprocessing.

[0027] Weighted fusion refers to the fusion of features of standardized data to obtain a comprehensive index that can fully characterize the dynamic characteristics of the processing process, namely, temperature-pressure-morphology weighted fusion data, which is conducive to subsequent analysis and closed-loop control. , which is used to characterize the dynamic characteristics of the machining process.

[0028] Step 3: Use FPGA chips and lightweight AI models to process standardized data and temperature-pressure-morphology weighted fusion data FdV in real time, dynamically optimize process parameters, and simultaneously generate temperature prediction and defect detection results, so that the subsequent steps can achieve low latency (≤10ms), high precision (error ±1°C) and energy efficiency improvement (energy consumption reduced by 40%) intelligent drilling and riveting control. The specific processing steps are as follows: Step b1: Build a lightweight AI model architecture: Based on the MobileViT model and convolutional neural network (CNN), the lightweight AI model's feature extraction capabilities for temperature fields and defect areas are improved through optimization technologies such as neural architecture search (NAS), knowledge distillation, and channel pruning.

[0029] Step b2: Hardware collaborative optimization: Split the lightweight AI model layer into parallel pipeline tasks, match the FPGA chip, and use the FPGA chip to cache pre-stored data (standardized data and temperature-pressure-morphology weighted fusion data FdV) to reduce external memory access latency.

[0030] Step b3: Improve the loss function: Combine the mean square error (MSE) and the structural similarity (SSIM) to improve the local detail retention capability of temperature prediction and balance the multi-objective conflicts. The loss function is improved to: ; The weight distribution (0.7:0.3) was verified through ablation experiments. The MSE weight of 0.7 can balance numerical accuracy and structural fidelity, reducing MSE by 20% and improving SSIM by 15%.

[0031] Step b4: Dynamically optimize process parameters: Update the temperature-pressure-morphology weighted fusion data FdV every 10 ms to respond to changes in processing status (such as temperature fluctuations, material deformation).

[0032] Step b5: Generate temperature prediction and defect detection results: Based on normalized temperature data And the normalized pressure data Predicted target temperature T opt ; Use CNN to binarize surface morphology image data I* Classify defect types (cracks, pores, deformation), output defect coordinates, types, severity levels, and provide real-time feedback to the parameter optimizer.

[0033] Step 4: Based on the process parameters, temperature prediction and defect detection results obtained in step 3, the Monte Carlo-NSGA-II hybrid algorithm is further optimized to generate candidate process parameters (temperature, pressure, displacement and speed), screen the Pareto frontier solution set (tensile strength ≥ 800MPa, energy consumption ≤ 500J / time, and the set of solutions that cannot further optimize the flow drilling and riveting performance by adjusting parameters without compromising other objectives), and transmit them to the digital twin case library in real time through the EtherCAT protocol. In this way, the digital twin drive parameter optimization is achieved.

[0034] Step 5: Construct a unified constitutive model of thermal-mechanical-phase change coupling, and use the multi-physics field coupling numerical analysis method to reversely optimize the Pareto frontier solution set. The Pareto frontier solution set is re-entered into the lightweight AI model through the digital twin case library for re-optimization to find better processing control process parameters; at the same time, the real-time monitored processing data is reversely input into the simulation model for closed-loop verification, realizing the virtual-actual closed-loop verification of the digital twin, which can reduce the error rate of heat affected zone (HAZ) morphology prediction by 30%.

[0035] Through multi-physics field coupling modeling and simulation, based on the phase change characteristics of titanium alloy in heat-assisted processing, a unified constitutive model of thermal-mechanical-phase change coupling is established, covering aspects such as high-temperature plasticity of materials, martensitic phase transformation dynamics and residual stress evolution, in order to reveal the drilling and riveting mechanism and facilitate the optimization of the drilling and riveting process. The unified constitutive model of thermal-mechanical-phase change coupling combines the Johnson-Cook plasticity model ( ) and the Avrami phase transition equation ( ), the specific equation is as follows: ; Where σ is the flow stress, A is the yield stress, B is the hardening modulus, , is the correction of the phase transformation volume fraction X to the initial yield stress (A0) and the initial hardening modulus (B0), , are the phase transformation softening coefficients of yield stress and hardening modulus, ε is the equivalent plastic strain, n is the hardening exponent, C is the strain rate sensitivity coefficient, is the normalized strain rate, is the normalized temperature, m is the temperature softening index, α is the phase change softening factor, X is the phase change volume fraction, k is the phase change rate, Equivalent plastic strain rate, t is time, ρ is density, is the specific heat at constant pressure, is the partial derivative of temperature with respect to time, Divergence, is the thermal conductivity, is the temperature gradient, η is the ratio of plastic work converted into heat, ΔH is the latent heat of phase change, is the phase change rate, is the residual stress change rate, E is the elastic modulus, is the coefficient of thermal expansion, is the temperature change, is the phase change expansion coefficient, is the change in volume fraction of phase change.

[0036] Step 6: The verified process parameters are included in the digital twin case library, which pushes new Pareto frontier solution sets to the lightweight AI model, thereby achieving dynamic updates. It has been proven that the new product process debugging cycle can be shortened by 40%.

[0037] The digital twin case library contains various cases and related information on process parameter optimization, such as simulation analysis, Pareto frontier solution sets, process parameter optimization data, and optimization control strategies.

[0038] Step 7: Dynamically modulate the PID parameters through the deep reinforcement learning model (DRL), combine with the verified process parameters, and control the heat-assisted flow drilling and riveting equipment through the PID control circuit (including driving the servo motor and heating module, etc., to achieve high-precision closed-loop control). Combined with the tensile strength-energy consumption balance reward function, the PID parameters are corrected in real time, and hardware-level protection is triggered when the limit is exceeded to ensure high processing precision and system stability, and the control results of the heat-assisted flow drilling and riveting equipment are fed back to the digital twin case library.

[0039] The state space of the deep reinforcement learning model (DRL) includes real-time processing data (temperature, pressure, displacement and surface topography image data) and digital twin input (heat affected zone (HAZ) topography prediction error, martensite ratio, residual stress, etc.). Its action space is divided into continuous actions (ΔK p (Proportional gain adjustment), ΔK i (Integral time constant adjustment), ΔK d (differential gain adjustment)) and discrete actions (path correction, exception handling).

[0040] The tensile strength-energy consumption balance reward function is: R=10×S 达标 +5×E 降低 −20×T 超限 −15×V 超标, where R is the reward value, S 达标 Indicates that the connection strength meets the standard, E 降低 Indicates reduced energy consumption, T 超限 Indicates temperature over limit mark; V 超标 Indicates that the vibration amplitude exceeds the limit.

[0041] The PID control circuit is as follows: Figure 3 As shown, it includes a proportional control part, an integral control part, a differential control part, and an adder circuit.

[0042] The proportional control part includes an operational amplifier and a resistor (R p ). Input signal (V in ) and the reference signal (V ref ) is input to the inverting input of the operational amplifier through a resistor network. Proportional gain (K p ) is composed of the feedback resistor (R f ) and the input resistance (R in ) ratio. Proportional gain determines the system response speed. Too large a proportional gain may cause oscillation, while too small a proportional gain may result in a slow response. Output signal (V p ) is the input deviation (V in -V ref ) and K p The product of .

[0043] The integral control part includes an operational amplifier, a capacitor and a resistor (R i ). Input signal (V in ) through a resistor (R i ) is input to the inverting input of the operational amplifier. A capacitor is connected to the feedback loop to accumulate the integral value of the input deviation. Integral gain (K i ) by the resistor (R i ) and capacitance. The integral gain can eliminate steady-state errors. Too large an integral gain may cause overshoot, while too small an integral gain may result in slow error elimination. Output signal (V i ) is the input deviation (V in -V ref ) and K i The product of .

[0044] The differential control part includes an operational amplifier, a capacitor, and a resistor (R d ). Input signal (V in ) is input to the inverting input of the operational amplifier through a capacitor. A resistor (R d ) is used to detect the rate of change of input deviation. Differential gain (K d ) by the resistor (R d) and capacitance. Differential gain can suppress system oscillation. Too large a gain may introduce noise, while too small a gain may not suppress the system oscillation. Output signal (V d ) is the input deviation (V in -V ref ) and K d The product of .

[0045] The adder circuit of the PID control module converts the ratio (V p )、Integral(V i ), Differential (V d ) The control output is input to the adder circuit through the resistor network. The adder linearly superimposes the three outputs to generate the total control signal (V out Total control signal (V out ) is used to drive the controlled object (such as servo motor, heating module, etc.). The total control signal (V out ) is calculated as follows: ; The dynamic adjustment process of the PID control is as follows: 1. Set initial parameters Digital twin pre-configuration: Provides initial parameters (K p =2.0, K i = 0.1, K d =0.5); Safety margin: Set the PID parameter variation range (K p ∈[1.0, 3.0], K i ∈[0.05, 0.15], K d ∈[0.3, 0.7]).

[0046] 2. Real-time control cycle steps (cycle ≤ 1ms), the steps are as follows: c1: read status data from the digital twin case library; c2: DRL policy network output ΔK p , ΔK i , ΔK d ; c3: Update PID parameters: K p + = K p +ΔK p , K i + = K i +ΔK i , K d + =K d +ΔKd ; c4: Calculate the control amount: ; c5: output control signal to the actuator; c6: Feedback actual processing data to the digital twin case library; c7: Update the DRL strategy network weights every 10 minutes to optimize the long-term control strategy.

[0047] 3. Establish an exception handling mechanism Over-limit threshold protection mechanism: When the monitored environmental parameters exceed the safety threshold (temperature ≥ 650°C or pressure ≥ 1500N), the "emergency power reduction" control strategy of deep reinforcement learning (DRL) is immediately triggered, and the hardware-level redundant protection circuit is activated simultaneously; Model failure fallback mechanism: When the PID output continues to deteriorate (such as the reward value is lower than the preset threshold for 10 consecutive times), it automatically switches to the conservative PID parameters preset by the digital twin.

[0048] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it and design structural methods and embodiments similar to the technical solution without creativity without departing from the purpose of the invention, they should all fall within the protection scope of the present invention.

Claims

1. A heat-assisted flow drilling and riveting control method based on digital twin, characterized in that: The following steps are involved: Step 1: Collect processing data of heat-assisted flow drilling and riveting equipment, including temperature, pressure, displacement and surface morphology image data of the processing area; Step 2: preprocess the temperature, pressure, displacement and surface morphology image data, standardize the preprocessed temperature, pressure and surface morphology image data, and then perform weighted fusion on the standardized temperature, pressure and surface morphology image data to obtain temperature-pressure-morphology weighted fusion data; Step 3: Use FPGA chips and lightweight AI models to process standardized data and temperature-pressure-morphology weighted fusion data in real time, dynamically optimize process parameters, and simultaneously generate temperature prediction and defect detection results; Step 4: Based on the process parameters, temperature prediction and defect detection results obtained in step 3, further optimization is performed through the Monte Carlo-NSGA-II hybrid algorithm to generate candidate process parameters, screen the Pareto frontier solution set, and transmit it to the digital twin case library in real time; Step 5: Construct a unified constitutive model of thermal-mechanical-phase change coupling, and use the multi-physics field coupling numerical analysis method to reversely optimize the Pareto frontier solution set. The Pareto frontier solution set is re-entered into the lightweight AI model through the digital twin case library for re-optimization, and the real-time monitored processing data is reversely input into the simulation model for closed-loop verification; Step 6: The verified process parameters are included in the digital twin case library, and the digital twin case library pushes a new Pareto frontier solution set to the lightweight AI model; Step 7: Dynamically modulate the PID parameters through the deep reinforcement learning model, combine with the verified process parameters, and control the heat-assisted flow drilling and riveting equipment through the PID control circuit.

2. The heat-assisted flow drilling and riveting control method according to claim 1, characterized in that: The temperature-pressure-morphology weighted fusion data , is the normalized temperature data, F * is the normalized pressure data, I * It is the surface topography image data after binarization.

3. The heat-assisted flow drilling and riveting control method according to claim 1, characterized in that: In step 7, when dynamically modulating the PID parameters through the deep reinforcement learning model, the PID parameters are corrected in real time in combination with the tensile strength-energy consumption balance reward function, and hardware-level protection is triggered when the limit is exceeded, and the control results of the heat-assisted flow drilling and riveting equipment are fed back to the digital twin case library.

4. The heat-assisted flow drilling and riveting control method according to claim 1, characterized in that: The state space of the deep reinforcement learning model includes real-time processing data and digital twin input, and the action space is divided into continuous actions and discrete actions.

5. The heat-assisted flow drilling and riveting control method according to claim 3, characterized in that: The tensile strength-energy consumption balance reward function is: R=10×S 达标 +5×E 降低 −20×T 超限 −15×V 超标 , where R is the reward value, S 达标 Indicates that the connection strength meets the standard, E 降低 Indicates reduced energy consumption, T 超限 Indicates temperature over limit mark; V 超标 Indicates that the vibration amplitude exceeds the limit.

6. The heat-assisted flow drilling and riveting control method according to claim 1, characterized in that: The process of step 3 is as follows: Step b1: Build a lightweight AI model architecture; Step b2: Split the lightweight AI model layer into parallel pipeline tasks, match the FPGA chip, and use the FPGA chip to cache pre-stored data; Step b3: The loss function is obtained by summing the mean square error and the structural similarity according to the weights; Step b4: dynamically optimizing process parameters; Step b5: Generate temperature prediction and defect detection results.

7. The heat-assisted flow drilling and riveting control method according to claim 6, characterized in that: In step b5, according to the normalized temperature data And the normalized pressure data Predicted target temperature T opt ; Use lightweight AI model to binarize surface morphology image data I * Classify defect types, output defect coordinates, types, and severity levels, and provide real-time feedback to the parameter optimizer.

8. A heat-assisted flow drilling and riveting control system for implementing the assisted flow drilling and riveting control method according to any one of claims 1 to 7, characterized in that: include: Multi-source data acquisition terminal, used to capture temperature, pressure, displacement and surface topography image data of the processing area in real time; Intelligent decision-making module, which consists of a lightweight AI model, parameter optimizer, digital twin case library and digital twin model library. The digital twin model library is used to store the thermal-mechanical-phase change coupled unified constitutive model and the Monte Carlo-NSGA-II hybrid algorithm; The dynamic control module is composed of a safety monitoring module, a PID control module, and a servo motor and a heating module of the heat-assisted flow drilling and riveting equipment; the PID control module is composed of a PID control circuit and a deep reinforcement learning model, and the PID parameters are dynamically modulated through the deep reinforcement learning model.

9. The heat-assisted flow drilling and riveting control system according to claim 8, characterized in that: Multi-source data acquisition terminals include infrared thermal imagers, multi-axis force sensors, laser displacement meters and industrial cameras.

Images