Self-adaptive hybrid wave soldering system and control method

Abstract

The invention discloses a self-adaptive hybrid wave soldering system and a control method, and relates to the technical field of wave soldering, the system comprises a digital twin control system, a self-adaptive solder supply system and a predictive maintenance module; a multi-physics field coordination module is arranged, when high-density circuit board wave soldering real-time control is carried out, a dynamic mapping model of a solder thermodynamic state and a fluid mechanics parameter is established, a turbulence suppression parameter and a thermal compensation coefficient are synchronously optimized, the solder heat-flow coupling change trend can be matched in real time, and the high-density circuit board wave soldering real-time control is achieved. Meanwhile, by actively inhibiting turbulence intensity fluctuation of molten solder, the reliability of a welding joint face is guaranteed, the wetting angle deviation risk is reduced, the welding defect rate is reduced, a dynamic posture sensing unit is arranged, technological parameters are corrected in real time through a multi-axis linkage compensation mechanism, the capillary action failure of the solder in pin gaps is avoided, and the welding quality is improved. And the collaborative stability of mechanical movement in the welding process is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of wave soldering, and in particular to an adaptive hybrid wave soldering system and a control method. Background Art

[0002] Wave soldering refers to the process of jetting molten soft solder into a solder wave required by the design through an electric pump or electromagnetic pump. It can also be formed by injecting nitrogen into the solder pool, so that a printed circuit board pre-installed with components passes through the solder wave to achieve soft soldering of mechanical and electrical connections between the component terminals, pins and printed circuit board pads. Wave soldering is to allow the welding surface of the plug-in board to directly contact with high-temperature liquid tin to achieve the purpose of welding. The high-temperature liquid tin maintains an inclined surface, and a special device is used to form the liquid tin into a wave-like phenomenon, so it is called "wave soldering". Its main material is solder bar.

[0003] At present, due to the demand for multi-variety mixed-line manufacturing in the production process of electronic components, when performing real-time control of high-density wave soldering of circuit boards, the equipped solder dynamic adjustment system acts on the welding station, and cannot match the coupled changes of the solder thermodynamic state and fluid mechanics parameters in real time. When the turbulence intensity fluctuations in the welding area are not suppressed in time, the wetting angle deviation will be greater than 5°, and the reliability of the welding joint surface cannot be guaranteed; at the same time, when performing continuous transmission welding of substrates, it is impossible to detect in real time whether the coordinated state of the substrate transmission inclination angle and the wave peak oscillation frequency is abnormal, which will cause the capillary action of the solder in the pin gap to fail, and the process parameters cannot be dynamically compensated when bridging defects occur; when welding on mixed production lines, due to the difference in thermal capacity of different board-type components, multi-physical field step-by-step closed-loop adjustment cannot be achieved when controlling the solder thermal field distribution, resulting in thermal boundary layer modeling mismatch between the high-density area and the heat dissipation area, affecting the overall welding yield and solder utilization efficiency.

[0004] Therefore, an adaptive hybrid wave soldering system and control method are proposed to solve the above problems. Summary of the Invention

[0005] The main purpose of the present invention is to provide an adaptive hybrid wave soldering system and a control method to solve the problems raised in the above background.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an adaptive hybrid wave soldering system and control method, the method comprising: S1. Mode switching decision: Receive production instructions, synchronize equipment status data with production plan data through the digital twin, calculate optimal working mode parameters based on the multi-physics field coupling simulation engine, and trigger seamless switching between online and offline modes; S2, solder closed-loop control: real-time collection of tin pot liquid level data, combined with the pressure wave feedback signal to dynamically calculate the predicted value of solder slag generation, and close-loop control of solder supply by adjusting the peak speed; S3. Fault prediction analysis: Continuously collect motor bearing vibration spectrum signals, analyze vibration characteristics through convolutional neural network (CNN), and predict the equipment's remaining life (RUL) and nozzle clogging risk level; S4. Predictive maintenance execution: When the remaining life is lower than a dynamic threshold, a preventive maintenance work order is automatically generated. The dynamic threshold is optimized and adjusted in real time based on historical failure data.

[0007] Preferably, the operation of the multi-physics field coupling simulation engine in step S1 specifically includes: The fluid mechanics sub-model simulates the turbulent flow state of the solder in the wave crest. The Reynolds-averaged Navier-Stokes equation is used to calculate the solder flow velocity distribution. The wave crest area is divided by an unstructured tetrahedral grid to adapt to the complex geometric shape. The specific configuration is as follows: Boundary layer treatment: the first layer of grid height near the wall is 0.1 mm, 5 layers of boundary layer grid are set, and the growth factor is 1.2; Regional densification strategy: The wave crest impact area uses a maximum grid size of 0.5mm + local curvature adaptive densification, the solder reflow channel uses a maximum grid size of 1.2mm + boundary-driven densification, and the main area of ​​the tin pot uses a 3.0mm uniform grid; The solver parameters are: Discrete format: pressure-velocity coupling: SIMPLEC algorithm, momentum equation: second-order upwind scheme, turbulence equation: first-order upwind scheme; Convergence control criteria: continuity equation residual is less than 10 -4 , the residual of the momentum equation is less than 10 -3 , the residual of the turbulence equation is less than 10 -2 ; Iteration settings: Maximum number of iterations per time step: 20 times; Sub-relaxation factor configuration: pressure equation: 0.3, momentum equation: 0.5, turbulent kinetic energy equation: 0.6; The thermodynamic sub-model calculates the temperature gradient distribution in the welding area through the Fourier heat conduction equation and combines it with the infrared thermal imager to calibrate the temperature field in real time. The mechanical vibration sub-model processes the 10kHz sampled vibration data through a spectrum analysis algorithm to identify the equipment's resonant frequency points and generate frequency avoidance parameters.

[0008] Preferably, the implementation of the welding slag generation amount prediction model in step S2 is: The material coefficient k is calibrated through experiments and has a value range of 0.15-0.35. The specific value depends on the solder tin-silver-copper alloy ratio. A quantitative relationship between the amount of slag generated ΔQ, the pressure wave feedback amplitude A, and the peak speed f is established: ΔQ=k·A·f The tin furnace temperature was controlled at 250°C, the fixed laminar wave height of the wave crest was 8mm, nitrogen protection was used to control oxidation so that the ambient oxygen concentration was less than 100ppm, the robotic arm performed automatic slag scraping every 15 seconds, the fixed wave crest speed f = 1200rpm, and the pressure wave amplitude A was gradually increased from 100Pa to 500Pa. The experiment was repeated 10 times for each set of parameters, and the amount of slag generated ΔQ was recorded. The average ΔQ of the 10 experiments was taken to eliminate random fluctuations. The pressure wave feedback amplitude ΔP is collected at a frequency of 200 Hz and is obtained through a piezoelectric sensor installed on the side wall of the tin furnace; The dynamic calculation unit performs model calculations every 500ms and outputs speed adjustment instructions to the brushless DC motor.

[0009] Preferably, the structure of the convolutional neural network CNN in step S3 includes: Input layer: receives 2048-point vibration spectrum data at 100kHz; Convolutional layer: uses a 3×3 convolution kernel to extract bearing wear features and uses the ReLU activation function to enhance nonlinear expression capabilities; Pooling layer: performs a maximum pooling operation to reduce feature dimensions; Fully connected layer: Outputs three-level classification results: normal, warning, fault, and remaining life regression value.

[0010] Preferably, the optimization process of the dynamic threshold in step S4 is specifically as follows: The historical fault database stores more than 500 bearing failure cases and records the deviation between the RUL prediction value and the actual failure time. The sliding window algorithm uses a 72-hour window length and dynamically updates the threshold formula: T threshold =α·μ error +β·σ error Where T threshold is the dynamic threshold, μ error is the mean forecast error of the 72-hour window, σ error is the standard deviation of the prediction error, α=1.2 is the mean weight coefficient, and β=0.8 is the standard deviation weight coefficient.

[0011] Preferably, the triggering logic for mode switching in step S1 includes: Online mode activation conditions: order priority score greater than 8 points and equipment load rate greater than 70%, activate the fully automatic transmission track and preset parameter welding process; Offline mode activation conditions: equipment load rate is less than 70%, more than 3 PCB types are processed simultaneously, and the semi-manual operation interface is started, allowing the operator to manually adjust the welding parameters; When switching modes, process parameters are synchronized via the OPC UA protocol, with data packet transmission delays less than 100ms.

[0012] Preferably, the system comprises: Digital twin control system, including: OPC UA communication module, which uses industrial Ethernet interface to synchronize device status data and production plan data; Multi-physics coupling simulation engine, integrating fluid mechanics model, thermodynamics model and mechanical vibration model; Mode switching actuator drives the pneumatic valve group to switch between online and offline mode pipelines; Adaptive Solder Feed System: Includes: The laser displacement sensor is installed on the side wall of the tin furnace at a 45° inclination to monitor liquid level changes; A brushless DC motor connected to the crest generator impeller through a planetary gearbox; Dynamic calculation unit with built-in slag generation prediction algorithm and PID speed controller; Predictive maintenance module, including: A vibration signal acquisition device, including a 100kHz broadband piezoelectric accelerometer; Blockage risk warning device, integrating CNN processor and fluid mechanics simulation interface.

[0013] Preferably, the hardware implementation of the multi-physics field coupling simulation engine is: The fluid dynamics sub-model runs on an FPGA chip, solving the Navier-Stokes equations in real time with a calculation cycle of less than 10ms. The thermodynamics sub-model is connected to an infrared thermal imager and receives temperature field data through a PCIe interface. The mechanical vibration sub-model is connected to a vibration acceleration sensor, which is installed on the motor base with a sampling rate of 10kHz.

[0014] Preferably, the closed-loop control process of the adaptive solder supply system includes: The laser displacement sensor collects the liquid level H every 50ms; The dynamic calculation unit compares H with the deviation ΔH of the set liquid level and calculates the compensation speed based on the pressure wave feedback signal: Among them, ΔRPM is the speed adjustment of the peak generator, ΔH is the tin furnace liquid level deviation, K p is the proportional gain coefficient, K d is the differential gain coefficient, is the instantaneous rate of change of liquid level deviation; The brushless DC motor performs speed regulation, and the impeller linear response time is less than 50ms.

[0015] Preferably, the early warning mechanism of the blockage risk early warning device is: The CNN processor performs vibration spectrum analysis every 10 minutes and outputs the wear level and RUL value; When the wear level is greater than the "warning" state, the fluid mechanics simulation interface is activated to simulate the shear stress distribution of solder flow: Where τ is the shear stress, μ is the dynamic viscosity, is the velocity gradient, u is the axial velocity of the fluid, y is the distance from the nozzle wall, is the partial differential operator; When the shear stress τ is greater than 0.35 Pa and the RUL is less than 120 hours, a level 3 warning work order is generated and the equipment is triggered to slow down.

[0016] The present invention has the following beneficial effects: 1. In the present invention, by setting up a multi-physics field coordination module, when performing real-time control of wave soldering of high-density circuit boards, by establishing a dynamic mapping model between the thermodynamic state of the solder and the fluid mechanics parameters, the turbulence suppression parameters and the thermal compensation coefficient are simultaneously optimized, and the trend of the thermal-fluid coupling change of the solder can be matched in real time; at the same time, by actively suppressing the fluctuation of the turbulence intensity of the molten solder, the reliability of the solder joint surface is guaranteed, the risk of wetting angle deviation is reduced, and the soldering defect rate is reduced.

[0017] 2. In the present invention, by setting a dynamic posture sensing unit, when performing continuous transmission welding of the substrate, the coordinated deviation value of the transmission inclination angle and the peak oscillation frequency is calculated in real time to determine whether the contact state between the substrate and the solder wave peak is abnormal; and when the bridge defect characteristics are detected, the process parameters are corrected in real time through the multi-axis linkage compensation mechanism to avoid the failure of the capillary action of the solder in the pin gap and ensure the coordinated stability of the mechanical movement during the welding process.

[0018] 3. In the present invention, by setting up a partitioned closed-loop execution system, when welding multiple types of panels on a mixed production line, independent control domains are divided according to the thermal capacity characteristics of different components, and differentiated control of different areas is achieved through a multi-physical field hierarchical adjustment strategy; the thermal boundary layer modeling correction parameters are output in real time to reduce the amount of solder oxide slag generated, thereby improving the overall welding yield and resource utilization efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a flow chart of an adaptive hybrid wave soldering control method of the present invention; Figure 2 This is a structural diagram of an adaptive hybrid wave soldering system of the present invention. DETAILED DESCRIPTION

[0020] With reference to the drawings and embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0021] Specific embodiments: An adaptive hybrid wave soldering system and control method, comprising the following steps: S1, mode switching decision: receiving production instructions, synchronizing equipment state data and production plan data through digital twin, calculating optimal working mode parameters based on multi-physical field coupling simulation engine, triggering seamless switching between online mode and offline mode; S2, solder closed-loop control: real-time acquisition of tin furnace liquid level data, dynamic calculation of solder slag generation amount prediction value combined with pressure wave feedback signal, closed-loop control of solder supply amount by adjusting wave peak speed; S3, fault prediction analysis: continuously collecting motor bearing vibration spectrum signals, analyzing vibration characteristics through convolutional neural network CNN, predicting device remaining useful life RUL and nozzle clogging risk level; S4, predictive maintenance execution: when the remaining useful life is lower than the dynamic threshold, automatically generating a preventive maintenance work order, and dynamically adjusting the threshold according to historical fault data in real time.

[0022] The operation of the multi-physical field coupling simulation engine in step S1 specifically includes: The fluid mechanics sub-model simulates the turbulent state of the solder in the wave peak, and the Reynolds averaged Navier-Stokes equation is used to calculate the solder flow velocity distribution, and the turbulent model equation is: Where k is the turbulent kinetic energy, ε is the turbulent dissipation rate, t is the time, u j is the velocity component of the fluid in the j direction, x j is the spatial coordinate in the j direction, v is the fluid kinematic viscosity, ν t =C μ k 2 / ε is the turbulent viscosity coefficient, σ k is the turbulent Prandtl number of k equation, P k is the generation term of turbulent kinetic energy, σ ε is the turbulent Prandtl number of ε equation, model constant C μ =0.09, σ k =1.0, σ ε =1.3, C ε1 =1.44, C ε2 =1.92, is the turbulent generation term; Implementation requirements: the boundary layer grid needs to meet less than 5 to accurately capture the near-wall turbulent characteristics; The thermodynamic sub-model calculates the temperature gradient distribution of the welding area through the Fourier heat conduction equation, and combines the real-time calibration of the temperature field by the infrared thermal imager; The mechanical vibration sub-model processes 10kHz sampling vibration data through a frequency spectrum analysis algorithm, identifies the device resonance frequency point, and generates frequency avoidance parameters.

[0023] The implementation of the slag generation amount prediction model in step S2 is as follows: The material coefficient k is calibrated by experiment, and the value range is 0.15-0.35. The specific value depends on the proportion of the solder tin-silver-copper alloy, and a quantitative relationship between the slag generation amount ΔQ and the pressure wave feedback amplitude A and the peak speed f is established: ΔQ=k·A·f Where the tin furnace temperature is controlled at 250±2℃, the fixed laminar wave height of the wave peak shape is 8mm, the environmental oxygen concentration is less than 100ppm by controlling oxidation with nitrogen protection, the mechanical arm performs automatic slag removal every 15 seconds, the fixed peak speed f=1200rpm, the pressure wave amplitude A gradually increases from 100Pa to 500Pa, each group of parameters is repeated 10 times, the slag generation amount ΔQ is recorded, the average value of ΔQ of 10 experiments is taken, and the random fluctuation is eliminated. The confidence interval formula of experimental data is: Where CI is the confidence interval, is the sample mean of the slag generation amount, s is the sample standard deviation, n=10 is the number of experimental repetitions, and t α / 2 is the t-distribution critical value, which is 2.262 when α=0.05; Implementation requirements: the confidence interval half-width of each experiment needs to be verified to be less than 5% of the average value, otherwise the number of experiments needs to be increased to n greater than 15; the collection frequency of the pressure wave feedback amplitude ΔP is 200Hz, which is obtained by the piezoelectric sensor installed on the side wall of the tin furnace; The dynamic calculation unit performs model operation every 500ms, and outputs the speed adjustment instruction to the brushless DC motor.

[0024] The structure of the convolutional neural network CNN in step S3 includes: Input layer: receiving 100kHz 2048-point vibration spectrum data; Convolution layer: using 3×3 convolution kernel to extract bearing wear features, and using ReLU activation function to enhance non-linear expression ability; Pooling layer: performing maximum pooling operation to reduce feature dimension; Fully connected layer: outputting three classification results: normal, warning, and fault, and remaining life regression value.

[0025] The optimization process of the dynamic threshold in step S4 is as follows: The historical fault database stores more than 500 bearing failure cases and records the deviation between the RUL prediction value and the actual failure time. The sliding window algorithm uses a 72-hour window length, and the threshold weight adaptive update formula is as follows: α t =α t-1 +η·(|μ error -μ actual |),β t =β t-1 -η·(|σ error -σ actual |) where μ actual is the mean error of actual failure time, μ error is the mean prediction error, σ actual is the actual standard deviation, σ error is the standard deviation of the prediction error, η=0.01 is the learning rate, α t is the mean weight coefficient at the current moment, α t-1 is the mean weight coefficient of the previous moment, β t is the standard deviation weight coefficient at time t, β t-1 is the standard deviation weight coefficient at time t-1; Implementation requirements: Weight coefficient α t , β t Updated every 24 hours, with the constraint range α∈[1.0,1.5], β∈[0.5,1.0]; dynamic update threshold formula: T threshold =α·μ error +β·σ error Where T threshold is the dynamic threshold, μ error is the mean forecast error of the 72-hour window, σ error is the standard deviation of the prediction error, α=1.2 is the mean weight coefficient, and β=0.8 is the standard deviation weight coefficient.

[0026] The triggering logic for mode switching in step S1 includes: Online mode activation conditions: order priority score greater than 8 points and equipment load rate greater than 70%, activate the fully automatic transmission track and preset parameter welding process; Offline mode activation conditions: equipment load rate is less than 70%, more than 3 PCB types are processed simultaneously, and the semi-manual operation interface is started, allowing the operator to manually adjust the welding parameters; When switching modes, process parameters are synchronized via the OPC UA protocol, with data packet transmission delays less than 100ms.

[0027] The system includes: Digital twin control system, including: OPC UA communication module, which uses industrial Ethernet interface to synchronize device status data and production plan data; Multi-physics coupling simulation engine, integrating fluid mechanics model, thermodynamics model and mechanical vibration model; Mode switching actuator drives the pneumatic valve group to switch between online and offline mode pipelines; Adaptive Solder Feed System: Includes: The laser displacement sensor is installed on the side wall of the tin furnace at a 45° inclination to monitor liquid level changes; A brushless DC motor connected to the crest generator impeller through a planetary gearbox; Dynamic calculation unit with built-in slag generation prediction algorithm and PID speed controller; Predictive maintenance module, including: A vibration signal acquisition device, including a 100kHz broadband piezoelectric accelerometer; Blockage risk warning device, integrating CNN processor and fluid mechanics simulation interface.

[0028] The hardware implementation of the multi-physics coupling simulation engine is as follows: The fluid dynamics sub-model runs on an FPGA chip, solving the Navier-Stokes equations in real time with a calculation cycle of less than 10ms. The thermodynamics sub-model is connected to an infrared thermal imager and receives temperature field data through a PCIe interface. The mechanical vibration sub-model is connected to a vibration acceleration sensor, which is installed on the motor base with a sampling rate of 10kHz.

[0029] The closed-loop control process of the adaptive solder supply system includes: The laser displacement sensor collects the liquid level H every 50ms; The dynamic calculation unit compares H with the deviation ΔH of the set liquid level and calculates the compensation speed based on the pressure wave feedback signal: Among them, ΔRPM is the speed adjustment of the peak generator, ΔH is the tin furnace liquid level deviation, K p is the proportional gain coefficient, K d is the differential gain coefficient, is the instantaneous rate of change of liquid level deviation; Integral anti-windup compensation term: Where ΔRPM integral is the integral term in the speed adjustment, K iis the integral gain coefficient, t-10s is 10 seconds before the current moment, t is the current moment, ΔH(τ) is the liquid level deviation changing with time τ, dτ is the time differential, e is the natural constant = 2.71828, λ = 0.2 is the attenuation factor, and the exponential term prevents integral saturation when the liquid level deviation is large; Implementation requirements: The total speed adjustment must meet ΔRPM total =ΔRPM+ΔRPM Integral , and |ΔRPM total |≤200rpm, where ΔRPM total is the total speed adjustment, ΔRPM is the basic adjustment, ΔRPM Integral It is the integral compensation adjustment amount; the brushless DC motor performs speed adjustment, and the impeller linear response time is less than 50ms.

[0030] The early warning mechanism of the blockage risk early warning device is: The CNN processor performs vibration spectrum analysis every 10 minutes and outputs the wear level and RUL value; When the wear level is greater than the "warning" state, the fluid mechanics simulation interface is activated to simulate the shear stress distribution of solder flow: Where τ is the shear stress, μ is the dynamic viscosity, is the velocity gradient, u is the axial velocity of the fluid, y is the distance from the nozzle wall, is the partial differential operator; When the shear stress τ is greater than 0.35Pa and the RUL is less than 120 hours, a level 3 warning work order is generated and the equipment is triggered to slow down. The speed reduction control law is: where f nominal is the rated peak speed, τ is the real-time shear stress, RUL is the remaining service life in hours, f new is the adjusted peak speed, and the cube root function ensures a smooth speed reduction in the late stage of wear; Implementation requirements: After deceleration, f new Greater than 800 rpm to avoid the risk of solder solidification.

[0031] Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations may be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. An adaptive hybrid wave soldering control method, characterized in that: The method comprises: S1. Mode switching decision: Receive production instructions, synchronize equipment status data with production plan data through the digital twin, calculate optimal working mode parameters based on the multi-physics field coupling simulation engine, and trigger seamless switching between online and offline modes; S2, solder closed-loop control: real-time collection of tin pot liquid level data, combined with the pressure wave feedback signal to dynamically calculate the predicted value of solder slag generation, and close-loop control of solder supply by adjusting the peak speed; S3. Fault prediction analysis: Continuously collect motor bearing vibration spectrum signals, analyze vibration characteristics through convolutional neural network (CNN), and predict the equipment's remaining life (RUL) and nozzle clogging risk level; S4. Predictive maintenance execution: When the remaining life is lower than a dynamic threshold, a preventive maintenance work order is automatically generated. The dynamic threshold is optimized and adjusted in real time based on historical failure data.

2. The adaptive hybrid wave soldering control method according to claim 1, characterized in that: The operations of the multi-physics field coupling simulation engine in step S1 specifically include: The fluid mechanics sub-model simulates the turbulent flow state of the solder in the wave crest and uses the Reynolds-averaged Navier-Stokes equation to calculate the solder flow velocity distribution; The thermodynamic sub-model calculates the temperature gradient distribution in the welding area through the Fourier heat conduction equation and combines it with the infrared thermal imager to calibrate the temperature field in real time. The mechanical vibration sub-model processes the 10kHz sampled vibration data through a spectrum analysis algorithm to identify the equipment's resonant frequency points and generate frequency avoidance parameters.

3. The adaptive hybrid wave soldering control method according to claim 1, characterized in that: The implementation method of the welding slag generation amount prediction model in step S2 is: The material coefficient k is calibrated through experiments and has a value range of 0.15-0.

35. The specific value depends on the solder tin-silver-copper alloy ratio. A quantitative relationship between the amount of slag generated ΔQ, the pressure wave feedback amplitude A, and the peak speed f is established: ΔQ=k·A·f The tin furnace temperature was controlled at 250±2°C, the fixed laminar wave height of the wave crest was 8mm, nitrogen protection was used to control oxidation so that the ambient oxygen concentration was less than 100ppm, the robotic arm performed automatic slag scraping every 15 seconds, the fixed peak speed f=1200rpm, and the pressure wave amplitude A was gradually increased from 100Pa to 500Pa. The experiment was repeated 10 times for each set of parameters, and the amount of slag generated ΔQ was recorded. The average ΔQ of the 10 experiments was taken to eliminate random fluctuations. The pressure wave feedback amplitude ΔP is collected at a frequency of 200 Hz and is obtained through a piezoelectric sensor installed on the side wall of the tin furnace; The dynamic calculation unit performs model calculations every 500ms and outputs speed adjustment instructions to the brushless DC motor.

4. The adaptive hybrid wave soldering control method according to claim 1, characterized in that: The structure of the convolutional neural network CNN in step S3 includes: Input layer: receives 2048-point vibration spectrum data at 100kHz; Convolutional layer: A 3×3 convolution kernel is used to extract bearing wear features, and the ReLU activation function is used to enhance nonlinear expression capabilities; Pooling layer: performs maximum pooling operation to reduce feature dimensions; Fully connected layer: Outputs three-level classification results: normal, warning, fault, and remaining life regression value.

5. The adaptive hybrid wave soldering control method according to claim 1, characterized in that: The optimization process of the dynamic threshold in step S4 is specifically as follows: The historical fault database stores more than 500 bearing failure cases and records the deviation between the RUL prediction value and the actual failure time; The sliding window algorithm uses a 72-hour window length and dynamically updates the threshold formula: T threshold =a·m error +b·s error Where T threshold is the dynamic threshold, μ error is the mean forecast error of the 72-hour window, σ error is the standard deviation of the prediction error, α=1.2 is the mean weight coefficient, and β=0.8 is the standard deviation weight coefficient.

6. The adaptive hybrid wave soldering control method according to claim 1, characterized in that: The triggering logic for mode switching in step S1 includes: Online mode activation conditions: order priority score greater than 8 points and equipment load rate greater than 70%, activate the fully automatic transmission track and preset parameter welding process; Offline mode activation conditions: equipment load rate is less than 70%, more than 3 PCB types are processed simultaneously, and the semi-manual operation interface is started, allowing the operator to manually adjust the welding parameters; When switching modes, process parameters are synchronized via the OPC UA protocol, with data packet transmission delays less than 100ms.

7. An adaptive hybrid wave soldering system, used to implement the adaptive hybrid wave soldering control method according to any one of claims 1 to 6, characterized in that: The system comprises: Digital twin control system, including: OPC UA communication module, which uses industrial Ethernet interface to synchronize device status data and production plan data; Multi-physics coupling simulation engine, integrating fluid mechanics model, thermodynamics model and mechanical vibration model; Mode switching actuator drives the pneumatic valve group to switch between online and offline mode pipelines; Adaptive Solder Feed System: Includes: The laser displacement sensor is installed at a 45° inclination on the side wall of the tin furnace to monitor liquid level changes; A brushless DC motor connected to the crest generator impeller through a planetary gearbox; Dynamic calculation unit with built-in slag generation prediction algorithm and PID speed controller; Predictive maintenance module, including: A vibration signal acquisition device, including a 100kHz broadband piezoelectric accelerometer; Blockage risk warning device, integrating CNN processor and fluid mechanics simulation interface.

8. The adaptive hybrid wave soldering system according to claim 7, characterized in that: The hardware implementation of the multi-physics field coupling simulation engine is as follows: The fluid mechanics sub-model runs on an FPGA chip, solving the Navier-Stokes equations in real time with a computational cycle of less than 10ms. The thermodynamic sub-model is connected to the infrared thermal imager and receives temperature field data through the PCIe interface; The mechanical vibration sub-model is connected to a vibration acceleration sensor, which is installed on the motor base with a sampling rate of 10kHz.

9. The adaptive hybrid wave soldering system according to claim 7, characterized in that: The closed-loop control process of the adaptive solder supply system includes: The laser displacement sensor collects the liquid level H every 50ms; The dynamic calculation unit compares H with the deviation ΔH of the set liquid level and calculates the compensation speed based on the pressure wave feedback signal: Among them, ΔRPM is the speed adjustment of the peak generator, ΔH is the tin furnace liquid level deviation, K p is the proportional gain coefficient, K d is the differential gain coefficient, is the instantaneous rate of change of liquid level deviation; The brushless DC motor performs speed regulation, and the impeller linear response time is less than 50ms.

10. The adaptive hybrid wave soldering system according to claim 7, characterized in that: The early warning mechanism of the blockage risk early warning device is: The CNN processor performs vibration spectrum analysis every 10 minutes and outputs the wear level and RUL value; When the wear level is greater than the "warning" state, the fluid mechanics simulation interface is activated to simulate the shear stress distribution of solder flow: Where τ is the shear stress, μ is the dynamic viscosity, is the velocity gradient, u is the axial velocity of the fluid, y is the distance from the nozzle wall, is the partial differential operator; When the shear stress τ is greater than 0.35 Pa and the RUL is less than 120 hours, a level 3 warning work order is generated and the equipment is triggered to slow down.

Images