Coating method for anti-static coating of LED circuit board

Abstract

The invention discloses a coating method of an anti-static coating of an LED circuit board, which comprises the following steps of: acquiring environment temperature and humidity, substrate surface temperature, roughness and main process parameters of spraying equipment in real time through a multi-source sensor, performing multi-stage denoising, drift correction and normalization processing on acquired data, and inputting the data into a pre-trained hydrodynamic prediction model; dynamically predicting the thickness distribution of the coating; by combining model prediction with target thickness comparison, parameter correction is calculated, the spraying pressure, the nozzle moving speed and the coating flow are automatically adjusted, and closed-loop control of the spraying process is achieved; meanwhile, thermal field control is reversely optimized in the curing process according to thickness distribution, multi-section temperature self-adaptive adjustment is achieved, the coating uniformity, automation and stability are remarkably improved, the online self-adaptive and anomaly detection capacity is achieved, and manual intervention and the defective product rate are effectively reduced.

Description

Technical Field

[0001] This invention relates to the field of automated surface coating technology, and in particular to a method for applying an antistatic coating to an LED circuit board. Background Technology

[0002] With the rapid development of the electronics manufacturing industry and the widespread application of LED circuit boards in various electronic products, surface coating for anti-static protection has become one of the key processes for improving product quality and reliability. Current anti-static coating processes for LED circuit boards generally employ mechanized spraying and traditional automated control methods to achieve uniform film coverage on the circuit board surface to prevent electrostatic damage. This process typically includes multi-station automated coating, spray atomization control, coating flow adjustment, and post-curing treatment, and its applications cover high-end LED displays, backlight modules, and automotive electronic circuit boards. Currently, antistatic coating spraying for LED circuit boards mainly employs PLC-based or industrial microcontroller unit (MCU)-based coating equipment. By setting fixed spraying parameters (such as working pressure, nozzle movement speed, and coating flow rate) and combining them with empirical values ​​based on ambient temperature and humidity, and substrate batches, the coating thickness and uniformity can be controlled. Some high-end production lines are equipped with basic ambient temperature and humidity detection and simple closed-loop feedback modules, enabling coarse self-adjustment of key parameters through preset program rules. Coating thickness typically relies on post-processing feedback, and in some cases, infrared thermography, laser profilometry, and other sensing methods are used for quality monitoring and correction of operating parameters. Furthermore, computational fluid dynamics (CFD) simulations and process parameter optimization algorithms have been introduced into some advanced production lines for offline optimization of the spraying process window. Currently, the main problems with existing technologies are: (1) Lack of real-time control capability of multi-variable coupling. Current spraying systems mostly rely on single parameter feedback or preset process windows, making it difficult to comprehensively adjust the dynamic changes of multi-variable coupling such as ambient temperature and humidity, substrate temperature, surface roughness, pressure, and speed in real time. As a result, the spraying quality is sensitive to working condition disturbances, and it is difficult to ensure the consistency and uniformity of coating thickness. (2) The online parameter adjustment accuracy is low. The existing feedback control is mostly a coarse rule adjustment, which cannot respond precisely to instantaneous changes in environment and equipment status. Especially when there are sudden changes in temperature and humidity or substrate temperature, equipment wear or paint batch replacement, the spraying parameter adjustment is delayed, which can easily lead to quality problems such as overspraying or underspraying, and local thickness deviation exceeding the standard. (3) It is impossible to realize intelligent closed loop based on multi-source sensor fusion and fluid dynamics prediction. Most systems rely only on local signals or empirical formulas and lack full-process multi-source data acquisition, fusion and deep coupling with physical models (such as fluid dynamics deposition behavior), which restricts the accurate adaptation and online prediction of process parameters under the changing spraying conditions. (4) The lack of an automatic parameter management mechanism for working condition switching makes it difficult for existing technologies to identify and classify typical spraying working condition changes (such as temperature, high humidity environment, paint batch replacement, etc.), and it is impossible to automatically call the corresponding optimal control strategy according to the working condition switching, resulting in weak system adaptability, frequent manual intervention, and affecting the intelligent upgrading of production lines and process yield. (5) Poor compatibility between curing thermal field and thickness distribution. Some systems do not fully consider the spatial differences in the thickness distribution of the previous spraying during the curing stage after coating, lack the ability to dynamically adjust the thermal field, which can easily lead to incomplete local curing or thermal stress concentration, affecting the final performance of the coating. Summary of the Invention

[0003] In order to solve the above-mentioned technical problems, the present invention provides a method for applying an antistatic coating to an LED circuit board.

[0004] The technical solution of this invention is implemented as follows: a method for applying an antistatic coating to an LED circuit board, comprising: S1: In the LED circuit board coating execution module, ambient temperature and humidity, substrate surface temperature, substrate roughness and spraying equipment operating parameters are collected. The spraying equipment operating parameters include spraying pressure, nozzle moving speed and paint flow rate. S2: Filter and normalize the collected multi-source sensor data to eliminate the impact of sensor drift and environmental interference on the accuracy of subsequent modeling. S3: Input the normalized sensor data into the pre-trained fluid dynamics prediction model to predict the deposition thickness distribution trend of the antistatic coating under the current working conditions. S4: Based on the predicted coating thickness distribution trend, calculate the deviation between the current spraying parameters and the target thickness, and generate a spraying parameter correction amount, which includes the spraying pressure adjustment value, the nozzle moving speed adjustment value, and the paint flow compensation coefficient. S5: Based on the spraying parameter correction amount, dynamically adjust the parameters of the spraying execution module to achieve closed-loop feedback control of spraying pressure, nozzle moving speed and paint flow rate; S6: In the curing control module, the thermal field distribution requirements are deduced from the coating thickness distribution, a temperature gradient control strategy matching the current coating distribution is generated, and the drying temperature curve is dynamically adjusted. S7: Collect the thickness detection data of the cured coating and compare it with the predicted thickness to generate a model error correction factor, which is used to update the input-output mapping relationship of the fluid dynamics prediction model. S8: Based on historical error correction factors and multi-condition operation data, construct a condition label mapping table to identify the typical condition category of the current spraying scenario; S9: When a change in working condition is detected, the system automatically switches to the corresponding spraying parameter adjustment strategy and thermal field control strategy to improve the system's adaptability under multivariable coupling conditions. S10: If the current coating thickness deviation is determined to exceed the preset threshold, an abnormal alarm mechanism is triggered, and the subsequent curing process is suspended, waiting for manual intervention or automatic parameter recalibration.

[0005] The method for applying an antistatic coating to an LED circuit board provided in this application has the following beneficial effects: (1) This invention integrates multiple sensors, including high-precision temperature and humidity sensors, infrared thermometers, and laser roughness sensors, into the coating execution module to collect real-time process data on the environment, substrate, and equipment operation. Through high-frequency acquisition and filtering correction at multiple spatial points and continuous time, signal noise and sensor drift are eliminated, ensuring high purity and consistency of input data, thus laying a solid data foundation for coating process control under complex working conditions. Compared with existing single-variable or low-frequency manual sampling methods, this invention significantly improves the response speed and global perception capability to parameter fluctuations, and significantly enhances the adaptability to dynamic working conditions. (2) This invention uses a multi-source data-driven CFD finite element prediction model to numerically predict the coating atomization, deposition and spreading behavior under the combined effects of temperature and humidity, substrate temperature, roughness and spraying process parameters. The model accurately describes the coating flow and adhesion through RANS equations and particle tracking algorithm, and outputs the thickness distribution trend in real time, ensuring high-quality film consistency; (3) Based on the model output and target thickness deviation, this invention adopts a weighted integral and PID / fuzzy-PI collaborative adjustment algorithm to adjust core process parameters such as spraying pressure, nozzle speed and paint flow rate in real time, so as to realize multivariate dynamic compensation under the physical quantity scale. Multivariate coupling and sliding mode control are carried out in parallel to suppress the diffusion of single-factor disturbances and ensure a smooth transition in the parameter switching process; (4) This invention generates a multi-segment PID temperature control strategy that is synchronized with the regional distribution by back-deriving the thermal field coupling model based on the thickness distribution, thereby realizing the intelligent allocation of curing oven power and airflow as needed. It automatically identifies the thermal hysteresis zone and implements thermal response compensation, effectively overcoming process defects such as insufficient curing or coking caused by uneven coating thickness, and significantly improving the overall product quality and the stability of subsequent processes. Attached Figure Description

[0006] Figure 1 This is a flowchart of an antistatic coating application method for LED circuit boards according to the present invention; Figure 2 This is a sub-flowchart of an antistatic coating application method for LED circuit boards according to the present invention; Figure 3This is another sub-flowchart of the method for applying an antistatic coating to an LED circuit board according to the present invention. Detailed Implementation

[0007] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0008] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0009] When used herein, the singular forms of “a,” “an,” and “the” may also include the plural forms unless the context clearly indicates otherwise. It should also be understood that the terms “comprising / including” or “having,” etc., specify the presence of the stated features, wholes, steps, operations, components, parts, or combinations thereof, but do not preclude the possibility of the presence or addition of one or more other features, wholes, steps, operations, components, parts, or combinations thereof. Meanwhile, the term “and / or” as used in this specification includes any and all combinations of the associated listed items.

[0010] Please see Figures 1-3 As shown, a method for applying an antistatic coating to an LED circuit board includes: S1: In the LED circuit board coating execution module, ambient temperature and humidity, substrate surface temperature, substrate roughness and spraying equipment operating parameters are collected. The spraying equipment operating parameters include spraying pressure, nozzle moving speed and paint flow rate. S2: Filter and normalize the collected multi-source sensor data to eliminate the impact of sensor drift and environmental interference on the accuracy of subsequent modeling. S3: Input the normalized sensor data into the pre-trained fluid dynamics prediction model to predict the deposition thickness distribution trend of the antistatic coating under the current working conditions. S4: Based on the predicted coating thickness distribution trend, calculate the deviation between the current spraying parameters and the target thickness, and generate a spraying parameter correction amount, which includes the spraying pressure adjustment value, the nozzle moving speed adjustment value, and the paint flow compensation coefficient. S5: Based on the spraying parameter correction amount, dynamically adjust the parameters of the spraying execution module to achieve closed-loop feedback control of spraying pressure, nozzle moving speed and paint flow rate; S6: In the curing control module, the thermal field distribution requirements are deduced from the coating thickness distribution, a temperature gradient control strategy matching the current coating distribution is generated, and the drying temperature curve is dynamically adjusted. S7: Collect the thickness detection data of the cured coating and compare it with the predicted thickness to generate a model error correction factor, which is used to update the input-output mapping relationship of the fluid dynamics prediction model. S8: Based on historical error correction factors and multi-condition operation data, construct a condition label mapping table to identify the typical condition category of the current spraying scenario; S9: When a change in working condition is detected, the system automatically switches to the corresponding spraying parameter adjustment strategy and thermal field control strategy to improve the system's adaptability under multivariable coupling conditions. S10: If the current coating thickness deviation is determined to exceed the preset threshold, an abnormal alarm mechanism is triggered, and the subsequent curing process is suspended, waiting for manual intervention or automatic parameter recalibration.

[0011] Step S1: In the LED circuit board coating execution module, ambient temperature and humidity, substrate surface temperature, substrate roughness, and spraying equipment operating parameters are collected. These spraying equipment operating parameters include spraying pressure, nozzle movement speed, and paint flow rate. Specifically, this includes: S1.1: Temperature and humidity data of the surrounding environment of the coating execution module are collected by temperature and humidity sensors to obtain environmental parameters that affect the volatilization and adhesion performance of the coating, which are used as one of the input conditions of the fluid dynamics model; Within the coating execution module, the surrounding space is used as the measurement object. A high-precision digital temperature and humidity sensor array is deployed to collect instantaneous values ​​of air temperature and relative humidity, thereby achieving real-time quantitative characterization of the process environment. Multi-point synchronous acquisition method (sampling period) ms, spatial point distribution ≥ (1), to achieve rapid capture of the spatial distribution of ambient temperature and humidity, and to aggregate the temperature and humidity data of each acquisition node to the data processing unit through a bus network, ensuring the spatiotemporal consistency of the input data; Furthermore, through digital signal preprocessing algorithms (including low-pass filtering and dynamic threshold removal, filter cutoff frequency...) Hz, removal threshold ± σ), suppressing abnormal fluctuations caused by sensor quantization noise and occasional interference, thereby obtaining stable effective values ​​of temperature and humidity; Furthermore, using the standard air state equation based on temperature With relative humidity Calculate the absolute humidity of the environment The formula is: in For temperature Lower saturated vapor pressure (Pa). The constant of water vapor (J / kg·K) The conversion factor for mass units is used for calculation. Used to characterize the water vapor content that affects the evaporation rate of coating solvents; Furthermore, based on the input requirements of the fluid dynamics model, the measured temperature and humidity values ​​and the calculated absolute humidity values ​​are combined to form an environmental parameter vector. They are labeled with timestamps and spatial node identifiers to form a structured measurement dataset; Through the above algorithm and processing method, the original environmental temperature and humidity signal obtained in the previous step is transformed into high-purity environmental characteristic data that can be directly input into the fluid dynamics model, so as to realize a quantitative description of the environmental dependence of coating volatilization and adhesion performance. For example, on an LED circuit board antistatic coating production line, sensor arrays are positioned at the four corners and the center of the spray booth, with a sampling period set to [missing information]. ms, data collected per second Group data; temperature measurement range ℃~ ℃, humidity measurement range %RH~ %RH. In one measurement, the temperature readings at the four corners were as follows: ℃ ℃ ℃ ℃, central position ℃, average humidity is %RH, after low-pass filtering and removal by three times the standard deviation, the standard deviations of both temperature and humidity are less than 1%. Calculate the corresponding points The value is approximately g / m³, fluctuation range less than g / m³. The processed data from the five nodes were combined to form a structured matrix and input into the CFD prediction module. The coating deposition rate error was reduced by approximately [missing value] compared to the unprocessed input. This significantly improves the stability of thickness control; S1.2: Non-contact measurement of the surface temperature of the LED circuit board substrate is performed using an infrared thermometer to obtain information on the current thermal state of the substrate, which serves as a key input parameter for predicting coating adhesion behavior. In the coating execution module, the surface of the LED circuit board substrate is used as the measurement object, and a high-precision infrared temperature measurement device (resolution) is deployed. ℃, response time (ms), enabling real-time non-contact measurement of substrate surface temperature; Multi-point scanning temperature measurement method (number of points ≥ (Several points were deployed to cover the central area and key edge areas) to obtain instantaneous temperature values ​​in different areas of the substrate, forming an initial data matrix of temperature spatial distribution; Furthermore, an optical non-uniformity correction algorithm is used (referencing the blackbody radiation reference model, temperature range). ℃~ (℃), correcting the measurement error caused by the difference in the response curve of the infrared detector, and obtaining the corrected temperature distribution data; Furthermore, based on the Stefan-Boltzmann law, an emissivity compensation model is used to correct the surface temperature value, where the surface temperature... The calculation formula is: in For the measured radiance value, The emissivity of the substrate surface. Stefan-Boltzmann constant ( W / m²·K 4 ); Furthermore, the Kalman filtering algorithm (process noise covariance matrix) is employed. Adaptive estimation smooths temperature measurements over time, reducing random fluctuations caused by equipment micro-vibrations and transient thermal disturbances; Furthermore, bilinear interpolation is performed on the smoothed multi-point temperature data to generate a continuous temperature field distribution map, which serves as an input to the fluid dynamics model to predict the spreading speed and curing trend of the coating under the current thermal state. Through the above measurement and processing link, the original infrared temperature measurement signal is transformed into a high-precision substrate temperature feature set that can be directly used for coating adhesion prediction, thereby achieving a precise quantitative description of the influence of thermal state on coating behavior. For example, on a certain production line, an infrared temperature measuring device is deployed at a distance from the spraying surface. At a distance of mm, the center point and two edge points are deployed in total. Temperature measurement location, temperature measurement range ℃~ ℃, emissivity calibration value set to In one measurement, the initial reading of the center point was... ℃, left edge is ℃, right edge is ℃, after correction for non-uniformity and emissivity, are respectively ℃ ℃ ℃, the instantaneous variance after Kalman filtering is reduced to The temperature difference range of the full-plate temperature field generated by interpolation is controlled within Within ℃. Inputting this temperature field into the CFD spraying prediction model reduces the simulation error of coating adhesion behavior by approximately [percentage missing] compared to the uncorrected temperature input. %, significantly improving coating thickness uniformity and deposition control stability; S1.3: A laser profilometer is used to scan the micro-morphology of the LED circuit board substrate surface and calculate the surface roughness Ra value to quantify the initial conditions of the substrate surface adsorption capacity and coating deposition characteristics. S1.4: Real-time pressure monitoring of the paint supply system of the spraying equipment is performed by pressure sensors to obtain spraying pressure data, which serves as one of the core process parameters affecting the atomization effect and deposition rate of the paint. S1.5: The nozzle movement speed is calculated based on the feedback signal of the servo motor encoder. Combined with the PLC control system, the set value and actual feedback value of the paint flow rate are obtained to form a closed-loop acquisition link for the spraying execution parameters, providing real-time basis for subsequent dynamic adjustment of parameters.

[0012] Step S2: Filtering and normalizing the acquired multi-source sensor data to eliminate the impact of sensor drift and environmental interference on the accuracy of subsequent modeling. Specifically, this includes: S2.1: Perform sliding window mean filtering on the collected ambient temperature and humidity, substrate surface temperature, substrate roughness, spraying pressure, nozzle moving speed and paint flow rate multi-source sensor data to suppress high-frequency noise interference in the sensor signals and obtain the raw sensor data after preliminary filtering. The raw multi-source sensor data collected in step S1 is used as the input, including ambient temperature and humidity, substrate surface temperature, substrate roughness and spraying equipment operating parameter signals. The acquired multi-source sensor data is first filtered using a sliding window mean filter method (parameter: window length). Sampling period This achieves smooth suppression of high-frequency random noise components, where the window length... Set according to the sensor response time and the expected smoothness; Furthermore, smoothed sequences are independently calculated for various sensor signals using sliding window mean filtering; Furthermore, by introducing a weight coefficient vector within the sliding window... A weighted moving average filter is used, where the weight coefficients are adaptively set according to the sensor signal-to-noise ratio and instantaneous dynamic response characteristics to enhance the ability to retain key numerical changes. Furthermore, a larger window length is set for slowly changing signals such as temperature, humidity, and pressure to improve smoothness, while a smaller window length is set for rapidly changing signals such as nozzle movement speed to maintain dynamic response. Furthermore, by utilizing boundary condition processing algorithms, the shortcomings in the initial stage of the sliding window are addressed. The data at points is mirror-filled or zero-filled to eliminate the impact of endpoint effects on data smoothness; By using sliding window mean and weighted mean filtering, the input multi-source sensor raw signals are transformed into pre-filtered raw sensor data, which effectively reduces the impact of high-frequency noise and instantaneous interference, providing a stable and reliable input basis for the subsequent wavelet threshold denoising in S2.2. For example, on an LED circuit board coating production line, the sampling period of the ambient temperature sensor is set to... ms, window length = The maximum fluctuation amplitude of the original temperature sequence within 10 seconds is ℃. Using the equal-weighted moving average formula, the fluctuation amplitude is reduced to ℃; In pressure sensor signal processing, the sampling period is ms, window length = The weighting coefficients are set to [0.2, 0.3, 0.5], and the highest instantaneous noise amplitude of the original pressure signal is... MPa, reduced to after filtering MPa. After image filling correction of the boundary segment signal, the continuity and smoothness of the data curve at the initial time are improved, and the average denoising efficiency of the subsequent wavelet denoising algorithm is improved by approximately The mean square error of the thickness prediction CFD model under the same working conditions is reduced by % compared to the unfiltered input. %; S2.2: Based on the sensor data after sliding window mean filtering, wavelet threshold denoising algorithm is used to perform time-frequency domain joint denoising on each sensor signal, extract the low-frequency trend component of each parameter, remove non-stationary noise components caused by equipment vibration or electromagnetic interference, and obtain the denoised sensor signal sequence. S2.3: Perform zero-point drift correction processing on each sensor signal after noise reduction. Based on the sensor reference value under historical no-load conditions, calculate the current signal offset and perform baseline correction on the current sensor data to obtain reference-aligned sensor data that eliminates the influence of zero drift. S2.4: Based on the reference-aligned sensor data, the minimum-maximum normalization method is used to perform interval mapping processing on each sensor parameter, and sensor data of different dimensions are uniformly mapped to the [0,1] normalized interval to generate normalized sensor feature vectors. S2.5: Perform correlation analysis and redundancy detection on the normalized sensing feature vectors, identify highly correlated parameter pairs based on the Pearson correlation coefficient matrix, remove redundant input features, and generate an optimized standardized sensing dataset for use as input to the subsequent fluid dynamics model.

[0013] Step S3: Input the normalized sensor data into the pre-trained fluid dynamics prediction model to predict the deposition thickness distribution trend of the antistatic coating under the current operating conditions. Figure 2 As shown, it specifically includes: S3.1: Perform feature encoding on the optimized ambient temperature and humidity, substrate surface temperature, substrate roughness, spraying pressure, nozzle moving speed and paint flow rate data to generate a multi-dimensional working condition vector that conforms to the input format of the fluid dynamics prediction model; S3.2: Based on the pre-trained computational fluid dynamics (CFD) finite element model, boundary conditions are mapped onto the multidimensional working condition vector to construct a dynamic simulation environment for the atomization, deposition and spreading behavior of the coating during the spraying process; Based on the multi-dimensional working condition vector generated from step S3.1, a boundary condition mapping algorithm (parameters: working condition vector dimension m, model meshing accuracy h) is used to realize the physical boundary condition loading and matching processing of the pre-trained CFD finite element model. Furthermore, through a spatial coordinate reconstruction method (parameter: nozzle scanning path curve function) (N) of discrete elements in the spraying area, to project the instantaneous position, velocity and spraying pressure parameters of the nozzle in the multi-dimensional working condition vector to the set of boundary nodes in the three-dimensional finite element calculation domain, and generate a node-level constraint condition matrix; Furthermore, a physical property parameter matching algorithm is adopted (parameter: dynamic viscosity of coating). Surface tension ,density This allows the process parameters such as temperature and humidity, coating flow rate and substrate temperature in the multi-dimensional working condition vector to be mapped to the physical property inputs in the fluid dynamics constitutive model, and then assigned to the attribute matrix of each element in the finite element mesh according to the node distribution. Furthermore, by using the initial boundary condition assembly method, the velocity inlet condition from the nozzle exit is... Atmospheric side constant pressure outlet conditions and the slip-free wall conditions on the substrate surface The boundary condition set B is uniformly transformed into CFD solution, forming a complete initial input framework for numerical simulation. Furthermore, a fusion technique of computational domain discretization and boundary condition interpolation is adopted (parameter: time step). The number of iterations (k_max) is used to smoothly interpolate the continuous working condition parameter field according to the finite element mesh distribution, while ensuring that the physical meaning of the original working condition is strictly preserved, so as to generate a dynamic simulation environment input dataset that can be used for subsequent CFD time-step iteration solution. The above method transforms the multidimensional working condition vector into the boundary conditions of the dynamic simulation environment acceptable to the pre-trained CFD finite element model, providing input conditions with physical consistency and spatial resolution matching for the numerical solution of paint atomization, deposition and spreading behavior in the subsequent spraying process, and achieving high-precision repeatability of deposition thickness trend prediction under different spraying conditions. For example, in an LED circuit board with a width of mm, length is In a mm-scale spraying scenario, the working condition vector dimension is set to... Including ambient temperature ℃, relative humidity % , substrate temperature ℃, spraying pressure MPa, nozzle speed m / s, paint flow rate ml / s. The number of elements in the finite element computational domain is set to [value missing]. Time step s, number of iterations The nozzle exit velocity boundary condition is set to... m / s, the spraying trajectory curve function was obtained after spatial coordinate reconstruction. This ensures the spray pattern completely covers the panel surface. After temperature and humidity correction, the dynamic viscosity of the coating is [value missing]. Pa·s, surface tension N / m, density kg / m 2 The aforementioned physical properties are loaded into the finite element attribute matrix, and a boundary condition set B is generated in conjunction with the no-slip wall condition to complete the simulation domain initialization. After performing boundary condition interpolation and fusion, a complete dynamic simulation environment input is generated. The deposition thickness trend map calculated by the CFD model under this condition is compared with the measured data, and the mean square error is [missing value]. The prediction accuracy meets the process control requirements; S3.3: In the dynamic simulation environment, the Reynolds-averaged Navier-Stokes equations are used to numerically solve the trajectory of paint particles in the airflow field to obtain the spatial distribution function of the paint deposition rate during the spraying process. Based on the dynamic simulation environment boundary condition input dataset constructed in step S3.2, the Reynolds-averaged Navier-Stokes (RANS) equations are solved using the method (parameters: turbulence model k-ε, discrete second-order upwind scheme, iterative convergence residual threshold). This enables the simultaneous numerical solution of the instantaneous and time-averaged motion states of coating particles in an airflow field; Furthermore, by discretizing the RANS equations using the control volume method, the three-dimensional flow field computational domain is divided into structured hexahedral elements. The momentum equation, continuity equation, and equations for turbulent kinetic energy and turbulent dissipation rate in the horizontal and vertical directions are then time-stepped to generate the gas phase velocity field. With pressure field Spatiotemporal distribution data; Furthermore, a Lagrange particle tracking algorithm is employed (parameters: particle diameter distribution, particle initial velocity vector). The force equilibrium of the coating particles in the solved air field is integrated to obtain their position vector. With velocity vector trajectory data; Furthermore, based on the geometric intersection of the Lagrange particle trajectory and the substrate surface, the particle volume deposited per unit time in each grid cell is calculated, and then the volume is determined using the volume conservation equation. Obtain the spatial distribution function of deposition rate ,in To capture the particle volume on the cell surface, For time step; Furthermore, to improve the stability of sedimentation rate prediction, a time-moving average method is used for each time step. The values ​​are smoothed, with the parameter being the number of historical window steps. Output smooth deposition rate distribution field ; By coupling RANS solution with particle tracking algorithm, the dynamic simulation environment is transformed into an accurate spatial distribution database of coating particle deposition rate, enabling high-fidelity prediction of key deposition characteristics during the spraying process. For example, in the scenario of anti-static spraying of LED circuit boards, let the particle size distribution range of the nozzle outlet cross-section be... μm to μm, initial velocity of the particles is m / s, spraying working pressure is MPa, discretizing the computational domain into Units, time step s,k-ε turbulence model parameters are taken Initial turbulent kinetic energy Turbulent dissipation rate The deposition rate is calculated by integrating the intersection points of the particle trajectory and the plate surface. Spatial distribution, average value is μm / s, with a maximum difference not exceeding After smoothing, the spatial mean square error was reduced by more than 15%. This result, used as the substrate adhesion correction input in step S3.4, effectively improved the fit between the predicted thickness trend and the measured data. S3.4: Combine the substrate surface roughness and temperature parameters to perform surface adhesion correction on the spatial distribution function of the coating deposition rate, so as to generate a local coating thickness prediction matrix that takes into account the substrate interface characteristics. Spatial distribution function of coating deposition rate based on the output of step S3.3 and the smoothed sedimentation rate field A surface adhesion correction method was adopted (parameter: substrate roughness). Substrate surface temperature Coating physical property parameter set This allows for the correction of interface characteristics in the predicted deposition rate. Furthermore, through the roughness influence coefficient calculation model (based on multi-scale contact mechanics theory), the roughness influence coefficient is calculated... Mapped to the effective contact area correction factor of the interface The calculation formula is: in and The roughness sensitivity coefficient is based on the experimentally fitted data. This is a roughness reference value; Furthermore, through a temperature influence coefficient calculation model (based on a correction factor for the viscosity-temperature dependence of the coating), the... Mapped to interface temperature correction factor The calculation formula is: in For temperature sensitivity coefficient, Reference substrate temperature; Furthermore, the roughness correction factor With temperature correction factor Perform product fusion to form a comprehensive surface adhesion correction coefficient. : Furthermore, As a scale factor acting on the smoothing deposition rate field The equivalent deposition rate field considering the substrate interface characteristics is obtained. : Furthermore, regarding the equivalent deposition rate field Perform time integration, with the integration interval being the duration of the spraying process. The local coating thickness prediction matrix is ​​obtained. ; By correcting for roughness and temperature, the deposition rate result calculated in step S3.3 is transformed into a local coating thickness prediction matrix that better reflects the actual spraying-adhesion process, thus achieving accurate modeling of coating deposition behavior under different surface conditions. For example, in the LED circuit board spraying process, the substrate roughness = μm, reference roughness = μm, roughness sensitivity coefficient = , = Calculations yielded = Substrate surface temperature = ℃, reference temperature = ℃, temperature sensitivity coefficient = Calculations yielded = , fusion = The smooth deposition rate field is in the central region. μm / s, reduced to μm / s after correction μm / s, in the edge region by μm / s corrected to μm / s. Spraying time = The predicted thickness of the central region is calculated by integration as s. μm, the predicted thickness of the edge region is μm, with the error between the measured data and the subsequent S7 step controlled within ± Within a certain percentage, it achieves precise adaptation and correction for different substrate surface conditions; S3.5: Based on the local coating thickness prediction matrix, perform two-dimensional spatial interpolation and normalization operations to generate a continuous coating thickness distribution trend map, which serves as a reference input for subsequent closed-loop feedback control.

[0014] Step S4: Based on the predicted coating thickness distribution trend, calculate the deviation between the current spraying parameters and the target thickness, and generate a spraying parameter correction amount, which includes the spraying pressure adjustment value, the nozzle movement speed adjustment value, and the paint flow compensation coefficient. Figure 3 As shown, it specifically includes: S4.1: Compare the coating thickness distribution trend output by the fluid dynamics prediction model with the preset target thickness point by point, calculate the local thickness deviation value at each sampling location, and obtain the thickness error distribution matrix in the spatial dimension. Based on the continuous coating thickness distribution trend map generated in step S3.5 and the preset target thickness distribution data matrix, a spatial point-by-point error calculation method is adopted (parameter: total number of grid nodes). Two-dimensional coordinate index This enables node-by-node comparison between the predicted thickness and the target thickness; Furthermore, the local thickness deviation at each sampling location is calculated using the absolute difference calculation formula. : in To predict the element values ​​of the thickness matrix, The values ​​of the target thickness matrix elements; Furthermore, a matrix-based batch processing strategy is adopted (parameter: number of matrix rows and columns). , The local deviations of the nodes in the global computational domain are calculated in parallel to generate a two-dimensional thickness deviation matrix. And retain the spatial coordinate index for subsequent weighted integration processing; Furthermore, a threshold limiting method is used (parameter: maximum permissible deviation). To suppress local spike errors caused by single-point prediction anomalies, node values ​​exceeding the maximum permissible deviation are forcibly truncated. And record the node coordinates of the limiting operation for anomaly analysis; Furthermore, combining the thickness deviation gradient calculation method (parameter: spatial step size) ),right Perform bidirectional difference calculations to extract the local error rate of change matrix. This is used for subsequent nozzle motion trajectory correction strategies; Through the above sequential processing, the difference between the predicted thickness trend and the target thickness is quantified at the node level, generating a thickness error distribution matrix with spatial resolution, thereby achieving accurate positioning and quantification of coating quality deviations. For example, in a piece of size mm× In the LED circuit board spraying scenario with a thickness of mm, the predicted thickness matrix output by S3.5 Using a 5mm×5mm grid, a total of × common One prediction node; target thickness matrix The process design value is set at 95μm across the entire range. The deviation values ​​obtained through point-by-point calculations range from -6.5μm to +5.8μm, with the node exhibiting the largest absolute value being [the value is missing from the original text]. =6.5μm, exceeding the ±5μm tolerance range required by the process, after applying a limiting correction, the abnormal node deviation is truncated to 5μm, and the coordinates are recorded. mm, mm). Based on the limiting matrix, a double-difference gradient calculation is performed to obtain the central region. The mean value is 0.04 μm / mm, and the value in the edge area is 0.15 μm / mm, indicating that the thickness error variation rate is greater at the edge. This data will be used as a key reference in the S4.4 nozzle speed adjustment strategy to achieve subsequent improvement in coating uniformity. S4.2: Based on the thickness error distribution matrix, a weighted integral algorithm is used to calculate the global thickness deviation index. The weighted integral algorithm introduces a position weight factor and a time decay factor to reflect the differences in the influence of different regions on the overall coating uniformity. The thickness error distribution matrix generated based on step S4.1 A weighted integral algorithm is used (parameter: location weight factor). Set, time decay factor This enables a comprehensive spatial-temporal evaluation of the thickness deviation at each node. Using the location weight allocation method (parameter: spatial coordinate index) (Based on the regional importance ranking table), the node locations in different regions are mapped to corresponding weight coefficients. This enables the quantification of the differentiated impact on key functional areas and peripheral areas; Furthermore, a method is constructed using a time decay function (parameter: historical deviation time index). Current calculation time ), time decay factor Modeled as an exponential decay coefficient : in This is a constant time decay coefficient used to control the decay rate of historical deviations; Furthermore, the location weight factor and the time decay factor are multiplied and fused to generate a comprehensive weight coefficient. : The weighted integral formula is used (parameter: number of nodes in the entire domain). Time sampling number ), calculate global thickness deviation index : By normalizing the weighted integral results, Mapped to the [0,1] interval, a global thickness deviation index vector is generated that can be used for comparison with the control threshold; Through the above algorithm processing method, the spatial thickness deviation of each node is transformed into a global comprehensive deviation index that takes into account both regional importance and time factors, so as to realize the quantitative evaluation and control input support of the overall coating uniformity. For example, in a size of mm× In the LED circuit board spraying process with a resolution of mm, the grid is divided into 30×60 nodes, and the position weight factor sets the weight of the central area of ​​the board surface (10×20 grid) as 1. The edge region is set as Time decay coefficient constant Set as s -1 ,calculate The range is between 0.95 and 0.60. The global local deviation value ranges from... to μm is obtained by weighted integration of comprehensive weights. = μm, the normalized result is This index is used to compare with the control threshold of 0.5 in step S4.3, triggering a spraying pressure gain compensation strategy to improve the coating thickness uniformity by more than 4.5%. This verifies the effectiveness and controllability of the weighted integral method combining position weight and time decay in global deviation calculation. S4.3: Based on the comparison between the global thickness deviation index and the preset control threshold, a spraying pressure adjustment value is generated. The spraying pressure adjustment value is calculated based on the PID regulation algorithm to achieve dynamic compensation of the fluid output capacity of the spraying system. S4.4: Dynamically plan and adjust the nozzle moving speed. Based on the spatial gradient characteristics of the thickness error distribution matrix, a fuzzy control strategy is used to generate the nozzle moving speed adjustment value to match the dynamic balance between the coating deposition rate and the substrate moving speed. S4.5: Based on the nonlinear coupling relationship between paint flow rate and spraying pressure, a compensation function model is constructed. According to the spraying pressure adjustment value and the nozzle moving speed adjustment value, the paint flow rate compensation coefficient is calculated to ensure that the spraying amount per unit area is maintained within the set range. S4.6: The spraying pressure adjustment value, nozzle movement speed adjustment value and paint flow compensation coefficient are fused together to generate a comprehensive spraying parameter correction value, which serves as the input command for the next stage of closed-loop feedback control.

[0015] Step S5: Based on the spraying parameter correction amount, dynamically adjust the parameters of the spraying execution module to achieve closed-loop feedback control of spraying pressure, nozzle movement speed, and paint flow rate. Specifically, this includes: S5.1: Based on the calculated spraying pressure adjustment value, the opening of the pneumatic regulating valve is adjusted by PID to obtain an output pressure value that matches the target pressure, ensuring stable spraying atomization effect; S5.2: Based on the nozzle movement speed adjustment value, the drive frequency of the servo motor is adjusted in a closed loop to obtain a nozzle scanning speed that matches the current working conditions and maintain the consistency of the coating deposition per unit area. Based on the nozzle movement speed adjustment value and current operating parameters, a servo-driven closed-loop control method is adopted (parameter: speed setpoint). Position encoder feedback value This enables real-time matching and adjustment of the nozzle scanning speed; Furthermore, through a gain-adaptive PID algorithm (parameter: proportional gain) Integral gain Differential gain Sampling period ) Calculate speed error Generate initial drive frequency adjustment amount ; Furthermore, a speed smoothing filter method is employed (parameter: sliding window width). ),right Moving average processing is performed to suppress frequency fluctuations caused by instantaneous velocity disturbances, resulting in a smoothed driving frequency adjustment amount. ; Furthermore, based on the motor-mechanical transmission coupling model (parameter: reduction ratio) Maximum speed ),Will Mapped to servo driver PWM output frequency adjustment value This ensures that the change in nozzle movement speed maintains a linear relationship with the desired coating deposition rate; Furthermore, a limiting and jerk constraint control method is employed (parameter: upper limit of frequency variation). accelerometer limit This suppresses mechanical shock and system vibration during the drive frequency change process, improving operational stability; Through a closed-loop feedback mechanism, the nozzle position change rate collected by the real-time position encoder is... and Continuous comparisons are made to bring the nozzle scanning speed error to within the allowable range, thereby achieving the goal of consistent coating deposition per unit area. For example, on an LED circuit board coating production line, the nozzle target movement speed = mm / s, the position encoder sampling resolution is mm, sampling period = ms. Actual measured feedback speed. = mm / s, then the speed error = mm / s. Select proportional gain. = Hz / (mm / s), Integral Gain = Hz / (mm / s·s), Differential gain = Hz·s / (mm / s), PID output = × + ×( × )+ ×( )= + + = Hz, after smoothing by a sliding window = Hz, mapped to reduction ratio = The servo driver PWM frequency increase value is Hz, after meeting the limiting condition, is output to the servo driver. After implementing this adjustment strategy, the nozzle movement speed stabilizes at Hz. ~ Within the range of mm / s, the consistency rate of coating deposition per unit area is improved to 99.2%; S5.3: Based on the paint flow compensation coefficient, the output pulse width of the metering pump is proportionally adjusted to achieve precise control of the paint supply per unit time and ensure the flow stability during the spraying process. Based on the paint flow compensation coefficient and the real-time operating status input data of the spraying execution module, a PWM pulse width proportional adjustment method is adopted (parameter: current PWM duty cycle). Target flow rate corresponding to duty cycle Flow compensation coefficient This method achieves the prediction and correction of the pulse width of the metering pump output. In this method, the target coating flow rate is... The product of the flow compensation coefficient and the target duty cycle is mapped to the flow rate compensation coefficient. The formula is: in Based on the calibration curve The obtained uncompensated duty cycle; Furthermore, the flow-duty cycle calibration curve fitting method (parameter: logarithm of calibration points) is used. Least squares fitting order The characteristic curve of the metering pump is modeled as a polynomial fitting function. And combined with the feedback value from the flow sensor Calculate real-time flow error ; Furthermore, a closed-loop proportional-integral (PI) control algorithm is adopted (parameter: proportional gain). Integral gain Sampling period ),Will Mapped to pulse width correction and superimposed on Forming the adjusted pulse width value : Furthermore, a pulse width smoothing filter method is employed (parameter: sliding window width). ),right Perform moving average processing to obtain the smoothed pulse width instruction. To suppress PWM output jitter caused by sampling noise or instantaneous operating condition disturbances; Furthermore, based on the response time constant of the metering pump drive circuit Implement slope limiting algorithm (parameter: maximum pulse width change rate) ),right The amplitude of the change is physically constrained to prevent the drive load from experiencing rapid impact; Through the above control algorithm, the flow compensation coefficient is converted into a physically executable PWM pulse width adjustment, so as to realize high-precision control of the amount of paint supplied per unit time by the metering pump under dynamic conditions, and ensure the flow stability and coating thickness consistency during the spraying process. For example, on an LED circuit board coating production line, the current target flow rate of the coating... = ml / min, obtained by calibration curve fitting = (Duty cycle). When calculating the flow compensation coefficient in step S4.5. = At that time, the target pulse width = Real-time traffic feedback = ml / min, error = ml / min, proportional gain = (Duty cycle unit / (ml / min)), Integral gain = (Duty cycle unit / (ml / min·s)), sampling period = s, the sum of historical integrals is ml / min, to obtain = × + × = .then = via sliding window width = After smoothing = At the maximum pulse width change rate = Output limited to / s under the condition After implementing this pulse width adjustment strategy, the flow rate stabilized at... ± Within the ml / min range, the coating thickness consistency rate is improved to 99.3%, meeting the technical indicators for high-precision spraying control; S5.4: Perform multivariate coupling verification on the adjusted spraying pressure, nozzle movement speed and paint flow rate to generate a comprehensive control state evaluation value, which is used to determine whether the current closed-loop control has reached the set control precision; S5.5: The adjusted spraying parameters are fed back to the hydrodynamic prediction model to update the model input variables, forming a closed-loop control data feedback mechanism, which provides real-time feedback for subsequent coating thickness prediction.

[0016] Step S6: In the curing control module, the thermal field distribution requirements are deduced from the coating thickness distribution to generate a temperature gradient control strategy that matches the current coating distribution, and the drying temperature curve is dynamically adjusted. Specifically, this includes: S6.1: Spatial discretization is performed on the coating thickness distribution data to obtain the spatial thickness distribution matrix along the surface of the LED circuit board, which serves as the initial boundary condition input for heat conduction modeling; S6.2: Based on the finite element analysis method of heat transfer, a coating-substrate coupled heat conduction model is constructed. The spatial thickness distribution matrix and material thermal property parameters are input, and the temperature field evolution process of each region under the standard curing curve is calculated. Based on the finite element method of heat transfer (parameters: element type, mesh density, time step, convergence threshold), a coupled heat conduction model of coating-substrate is constructed to realize the temperature field simulation calculation of the curing process; Furthermore, by importing the spatial thickness distribution matrix and material thermal property parameters (parameter: coating thermal conductivity)... thermal conductivity of substrate ,density Specific heat capacity In the simulation domain, each mesh element is assigned the properties of a layered composite material, and heat flux continuity constraints are established for the interface nodes. Furthermore, the unsteady-state heat conduction control equations are discretized and solved, with the difference form being: in , , These are the unit material density, specific heat capacity, and thermal conductivity, respectively. For temperature, For time step; Furthermore, a standard curing curve is applied as the initial boundary condition (parameters: heating rate, isothermal time, cooling rate), and a convective heat transfer boundary is introduced in the top element of the model (parameter: convection coefficient). Ambient temperature ), to simulate real thermal field environments; Furthermore, the implicit Crank-Nicolson time integration scheme (parameter: time weighting coefficient) is utilized. = Numerical solutions are used to solve the discrete equation system, and convergence is achieved through iteration (condition: temperature increment norm ≤ convergence threshold). Obtain the temperature value of each grid cell within each time step; By using finite element numerical calculations, the temperature field evolution data under the action of the standard curing curve is obtained by mapping the coating-substrate model, thus realizing the simulation of the curing thermal process with high spatial resolution and strong temporal continuity. For example, in the LED circuit board curing simulation, the 0.12mm thick antistatic coating and the 1.6mm thick FR-4 substrate are discretized into octahedral units, and the coating thermal conductivity is... = W / (m·K), thermal conductivity of the substrate = W / (m·K), coating density = kg / m³, specific heat capacity = J / (kg·K), substrate density = kg / m³, specific heat capacity = J / (kg·K). A grid density of 0.5 mm and a time step of [missing information] were used. = s, convective heat transfer coefficient = W / (m²·K), ambient temperature = The curing curve is defined as a heating rate of 1.5 K / s to 150 °C, held at that temperature for 30 min, and then allowed to cool naturally. Simulation results show that the thermal response delay in the thick coating area is about 12 s, and the local temperature peak is 3.4 K lower than that in the thin coating area. This provides a basis for identifying the thermal hysteresis area and supports the generation of the thermal response difference index in the subsequent S6.3 step. S6.3: Based on the temperature field evolution data output by the heat conduction model, identify the local thermal hysteresis region caused by uneven coating thickness distribution, and generate a thermal response difference index matrix to quantify the curing response differences in different regions. S6.4: Based on the thermal response difference index matrix, a fuzzy clustering algorithm is used to classify the coating curing response region categories, forming a multi-segment thermal field control label map, providing a strategic basis for zoned temperature control; S6.5: Based on the multi-segment thermal field control label diagram, the preset temperature gradient strategy template is called to generate a dynamic drying temperature curve that matches the current coating distribution. The output power of each heating segment is controlled in real time through a PID controller to achieve adaptive matching of the thermal field distribution.

[0017] Step S7: Collect the thickness detection data of the cured coating and compare it with the predicted thickness to generate a model error correction factor, which is used to update the input-output mapping relationship of the fluid dynamics prediction model. Specifically, this includes: S7.1: Non-contact thickness measurement of the cured LED circuit board coating is performed. A laser thickness gauge is used to obtain the thickness values ​​of the coating surface at multiple sampling points to form a coating thickness distribution matrix, which serves as the benchmark data for model error analysis. S7.2: Spatially align and numerically match the coating thickness distribution matrix with the predicted thickness distribution output by the hydrodynamic prediction model, calculate the prediction error value of each sampling point, and construct a thickness error distribution map as a basis for model deviation quantification. S7.3: Based on the thickness error distribution map, perform error attribution analysis to identify the main error sources, including sensor measurement errors, model parameter drift and environmental disturbances, in order to generate error component decomposition vectors to guide the direction of model correction; Based on the thickness error distribution map, a multi-source data feature clustering analysis method (parameters: number of clusters k, similarity measurement method, iteration termination threshold) is used to achieve preliminary partitioning of error patterns and map different types of thickness error samples to error signal categories under specific working conditions. Furthermore, the prediction error signal is decomposed by using an error component separation algorithm (parameters: number of Gaussian mixture model components, number of expectation maximization iterations), separating the error signal into a measurement noise subset, a model parameter offset subset, and an external disturbance subset; Furthermore, for the subset of measurement noise, a sensor characteristic correction model is adopted (parameter: zero-point drift coefficient). proportional deviation coefficient Noise variance The sensor measurement error contribution rate is calculated by fitting the data. Furthermore, for a subset of model parameter offsets, residual time series analysis (parameters: autocorrelation order p, partial autocorrelation order q) is used to estimate the systematic offset between the model output and the actual thickness. And calculate its proportion of the total error; Furthermore, for a subset of external disturbances, correlation regression analysis of environmental variables was performed (parameter: significance level). Correlation threshold ) Calculate the regression coefficient vector of the error caused by temperature and humidity fluctuations, substrate temperature changes, etc. The contribution rate of external disturbances is obtained.

[0018] By normalizing the error contribution rate, the proportion of each error source is assembled into an error component decomposition vector, thereby achieving quantitative guidance for the direction of model correction. For example, in a certain batch of LED circuit board coating, the variance of the thickness error distribution obtained from laser thickness measurement statistics is: μm², cluster analysis was set to k=3, Euclidean distance was used as the similarity measure, and the three types of errors were attributed to: the proportion of sensor measurement noise. % (Zero-point drift coefficient) proportional deviation coefficient Noise variance μm²), proportion of model parameter offset % (systematic offset) μm), proportion of environmental disturbance (Temperature correlation coefficient) Humidity correlation coefficient Substrate velocity correlation coefficient After normalization, the error component decomposition vector is formed. As the input weight factor for the subsequent weighted least squares model update in S7.4, it effectively improves the targeting and convergence speed of model correction; S7.4: Based on the error component decomposition vector and the historical error correction factor, the weighted least squares method is used to calculate the mapping deviation between the current model output and the target value, and a model parameter update vector is generated to correct the input-output mapping relationship of the fluid dynamics prediction model. S7.5: Load the model parameter update vector into the parameter matrix of the fluid dynamics prediction model, and perform online model parameter update operation to improve the prediction accuracy and stability of the model under the current working conditions, forming a closed-loop learning mechanism.

[0019] Step S8: Based on historical error correction factors and multi-condition operating data, construct a condition label mapping table to identify the typical condition category of the current spraying scenario. Specifically, this includes: S8.1: Perform cluster analysis on historical error correction factors to identify the distribution pattern of model prediction errors under different working conditions and obtain multiple error feature clusters as the initial label basis for working condition classification. S8.2: Feature extraction and dimensionality reduction are performed on the ambient temperature and humidity, substrate surface temperature, substrate roughness, spraying pressure, nozzle moving speed and paint flow rate in the multi-condition operation data. The principal component analysis algorithm is used to extract the feature vectors of key process parameters to form the condition feature space. S8.3: Perform joint mapping analysis between the error feature cluster and the operating condition feature space, establish the correlation between error features and process parameter features based on the K-nearest neighbor algorithm, and generate preliminary operating condition label mapping rules; After obtaining the error feature cluster from S8.1 output and the operating condition feature space from S8.2 output, a joint mapping analysis method is used (parameters: Euclidean distance metric, feature weight coefficient). (Feature normalization interval), to achieve feature alignment and matching across different data domains; Furthermore, for each error feature cluster center vector With the sample vector of the working condition feature space Using the K-nearest neighbors algorithm (parameter: neighborhood size) Distance-weighted index ) Calculate the weighted Euclidean distance between features to achieve a similarity measurement between cross-domain errors and process features; Furthermore, all operating condition feature samples are sorted in ascending order of distance value, and the K nearest neighbor sample sets are selected. And count the number of times the same working condition label appears in the set to form a neighbor label frequency histogram; Furthermore, based on the frequency distribution of neighbor labels, a comprehensive score for each label is calculated using a distance-inverse weighting strategy. : Furthermore, the label with the highest comprehensive score is selected as the association label between the current error feature cluster and the working condition feature space, thereby generating the initial rules for label allocation; Through the above K-nearest neighbor association analysis processing method, the cross-domain error pattern is accurately bound to the specific process feature, and a preliminary working condition label mapping rule containing working condition label, corresponding feature range and error pattern description is generated, so as to achieve the technical effect of subsequent rapid working condition identification and strategy adaptation. For example, in the coating production of a certain batch of LED circuit boards, the number of error feature clusters output by S8.1 is 3, and the dimension of the center vector is 3, corresponding to sensor noise-dominated type [0.82, 0.11, 0.07], model offset-dominated type [0.15, 0.74, 0.11], and environmental disturbance-dominated type [0.21, 0.14, 0.65], respectively; S8.2 obtains 120 samples in the working condition feature space, with a feature dimension of 6. After normalization, the minimum weighted Euclidean distance between the center of the sensor noise-dominated feature cluster and a certain sample set is Neighborhood size Within this neighborhood, the operating condition label statistics show that "low temperature and high humidity" appeared 3 times and "normal temperature and medium humidity" appeared 2 times. After weighted calculation using the inverse distance, the overall score for "low temperature and high humidity" was [score missing]. The overall score for "normal temperature and moderate humidity" is: Therefore, this error feature cluster was associated and labeled as "low temperature and high humidity sensor noise enhancement condition", which achieved a high confidence match between specific error patterns and specific process conditions, providing an accurate label basis for subsequent decision tree optimization and real-time identification; S8.4: Perform decision tree optimization on the working condition label mapping rules to improve the interpretability and generalization ability of working condition classification, and output the optimized working condition label mapping table for real-time working condition identification and strategy matching. S8.5: Based on the currently collected multi-source sensor data and the output of the fluid dynamics prediction model, calculate the feature vector of the current spraying scene in real time, match and compare it with the working condition label mapping table, identify the typical working condition category to which the current spraying scene belongs, and output the working condition identification result.

[0020] Step S9: When a change in operating condition is detected, automatically switch to the corresponding spraying parameter adjustment strategy and thermal field control strategy to improve the system's adaptability under multivariable coupling conditions. Specifically, this includes: S9.1: Based on historical error correction factors and multi-condition operation data, construct a condition label mapping table. The condition labels include ambient temperature and humidity range, substrate surface temperature level, substrate surface roughness range, spraying pressure fluctuation range, and nozzle movement speed level, so as to realize a multi-dimensional feature description of the current spraying scenario. S9.2: Perform feature normalization processing on the currently collected ambient temperature and humidity, substrate temperature, substrate roughness, spraying pressure and nozzle moving speed to eliminate the influence of the difference in the dimensions of different sensors on the accuracy of working condition identification and obtain a standardized working condition feature vector. S9.3: The K-nearest neighbor classification algorithm is used to perform pattern recognition on the standardized working condition feature vector. Based on the working condition label mapping table, the typical working condition category to which the current spraying scenario belongs is matched to determine the current process environment status of the system. S9.4: Based on the identified typical working condition categories, call the preset spraying parameter adjustment strategy and thermal field control strategy template to generate spraying pressure adjustment value, nozzle movement speed adjustment value and paint flow compensation coefficient that match the current working condition, so as to optimize coating deposition behavior. S9.5: Based on the thermal field control strategy template corresponding to the current working condition category, generate a temperature gradient control curve that matches the coating thickness distribution, and dynamically adjust the heating power output and wind speed distribution of the curing module to improve the coating curing uniformity and process stability. S9.6: During the switching of operating conditions, a sliding mode control strategy is adopted to gradually adjust the spraying parameters and thermal field control parameters to suppress system disturbances caused by the switching of control strategies and improve the dynamic response stability of the system under multivariable coupling conditions.

[0021] Step S10: If the current coating thickness deviation exceeds a preset threshold, an abnormal alarm mechanism is triggered, and the subsequent curing process is paused, awaiting manual intervention or automatic parameter recalibration. Specifically, this includes: S10.1: Compare the actual coating thickness data collected after curing with the target thickness value, calculate the absolute value of the thickness deviation to quantify the execution error of the current spraying control module, and output the thickness deviation index. ; S10.2: Based on thickness deviation index Compared with the preset process tolerance threshold Compare and judge, if > Then an abnormal trigger signal is generated. This serves as the input condition for initiating the exception handling process; S10.3: In response to an abnormal trigger signal The system activates the audible and visual alarm device and generates an alarm log to notify the operator that there are abnormal fluctuations in the current spraying control process, requiring manual inspection or system intervention. S10.4: Send a process pause command to the PLC control system to interrupt the heating and conveyor belt operation of the curing module, prevent substandard coatings from entering the curing stage, and avoid resource waste and defective products from flowing out. S10.5: Determine if an automatic calibration enable signal exists. If it exists, initiate the parameter recalibration process based on the historical error correction factor and the current multi-source sensor data to optimize the fluid dynamics prediction model and spraying control parameters. If it does not exist, maintain the paused state and wait for manual intervention instructions.

[0022] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0023] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and rules of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for applying an antistatic coating to an LED circuit board, characterized in that, Includes the following steps: S1: In the LED circuit board coating execution module, ambient temperature and humidity, substrate surface temperature, substrate roughness and spraying equipment operating parameters are collected. The spraying equipment operating parameters include spraying pressure, nozzle moving speed and paint flow rate. S2: Filter and normalize the collected multi-source sensor data to generate normalized sensor data; S3: Input the normalized sensing data into the pre-trained fluid dynamics prediction model to predict the deposition thickness distribution trend of the antistatic coating under the current working conditions. S4: Based on the predicted coating thickness distribution trend, calculate the deviation between the current spraying parameters and the target thickness, and generate the spraying parameter correction amount; S5: Dynamically adjust the parameters of the spraying execution module according to the spraying parameter correction amount; S6: In the curing control module, the thermal field distribution requirements are deduced from the coating thickness distribution, a temperature gradient control strategy matching the current coating distribution is generated, and the drying temperature curve is dynamically adjusted. S7: Collect the thickness detection data of the cured coating and compare it with the predicted thickness to generate a model error correction factor; S8: Based on historical error correction factors and multi-condition operation data, construct a condition label mapping table to identify the typical condition category of the current spraying scenario.

2. The method for applying an antistatic coating to an LED circuit board according to claim 1, characterized in that, Following step S8, the following is also included: S9: When a change in working condition is detected, automatically switch to the corresponding spraying parameter adjustment strategy and thermal field control strategy; S10: If the current coating thickness deviation is determined to exceed the preset threshold, an abnormal alarm mechanism is triggered, and the subsequent curing process is suspended, waiting for manual intervention or automatic parameter recalibration.

3. The method for applying an antistatic coating to an LED circuit board according to claim 1, characterized in that, Step S1 specifically includes: Temperature and humidity data of the surrounding environment of the coating execution module are collected by temperature and humidity sensors to obtain environmental parameters that affect the evaporation and adhesion performance of the coating. The surface temperature of the LED circuit board substrate is measured non-contactly using an infrared temperature measuring device to obtain the current thermal state information of the substrate. A laser profilometer was used to scan the microstructure of the LED circuit board substrate surface and calculate the surface roughness Ra value. Real-time pressure monitoring of the paint supply system of the spraying equipment is performed using pressure sensors to obtain spraying pressure data. The nozzle movement speed is calculated based on the feedback signal from the servo motor encoder. Combined with the PLC control system, the set value and actual feedback value of the paint flow rate are obtained to form a closed-loop acquisition link for the spraying execution parameters.

4. The method for applying an antistatic coating to an LED circuit board according to claim 3, characterized in that, The temperature and humidity data acquisition specifically involves deploying a high-precision digital temperature and humidity sensor array and employing multi-point synchronous acquisition, low-pass filtering, and dynamic threshold rejection algorithms to obtain high-purity environmental characteristic data.

5. The method for applying an antistatic coating to an LED circuit board according to claim 1, characterized in that, Step S2 specifically includes: The collected multi-source sensor data is subjected to sliding window mean filtering to obtain the raw sensor data after preliminary filtering. Based on the original sensor data after preliminary filtering, a wavelet threshold denoising algorithm is used to perform time-frequency domain joint denoising on each sensor signal to obtain a denoised sensor signal sequence. Zero-point drift correction is performed on each denoised sensor signal. Based on the sensor reference value under historical no-load conditions, the current signal offset is calculated, and baseline correction is performed on the current sensor data to obtain reference-aligned sensor data. Based on the sensor data aligned to the benchmark, the minimum-maximum normalization method is used to perform interval mapping on each sensor parameter to generate a normalized sensor feature vector. Correlation analysis and redundancy detection are performed on the normalized sensing feature vectors. Highly correlated parameter pairs are identified based on the Pearson correlation coefficient matrix, redundant input features are removed, and an optimized standardized sensing dataset is generated.

6. The method for applying an antistatic coating to an LED circuit board according to claim 1, characterized in that, Step S3 specifically includes: The optimized standardized sensor dataset is feature-encoded to generate a multi-dimensional working condition vector; Based on the pre-trained computational fluid dynamics finite element model, boundary conditions are mapped onto the multidimensional working condition vectors to construct a dynamic simulation environment for the atomization, deposition, and spreading behavior of the coating during the spraying process. In the dynamic simulation environment, the Reynolds-averaged Navier-Stokes equations are used to numerically solve the trajectory of paint particles in the airflow field, and the spatial distribution function of paint deposition rate during the spraying process is obtained. By combining the substrate surface roughness and temperature parameters, the spatial distribution function is corrected for surface adhesion to generate a local coating thickness prediction matrix. Based on the local coating thickness prediction matrix, two-dimensional spatial interpolation and normalization operations are performed to generate a continuous coating thickness distribution trend map.

7. The method for applying an antistatic coating to an LED circuit board according to claim 6, characterized in that, In step S3, the fluid dynamics prediction model is a pre-trained fluid dynamics finite element model. After inputting the multi-dimensional working condition vector, the model outputs the spatial distribution function of the coating deposition rate through boundary condition mapping, property matching and initial boundary value assembly. The model is then combined with the substrate surface roughness and temperature parameters to perform surface adhesion correction and time integration, thereby obtaining the local coating thickness prediction matrix.

8. The method for applying an antistatic coating to an LED circuit board according to claim 1, characterized in that, Step S4 specifically includes: The coating thickness distribution trend output by the fluid dynamics prediction model is compared with the preset target thickness point by point, and the local thickness deviation value at each sampling position is calculated to obtain the thickness error distribution matrix in the spatial dimension. Based on the thickness error distribution matrix, the global thickness deviation index is calculated; Based on the comparison result between the global thickness deviation index and the preset control threshold, a spraying pressure adjustment value is generated; The nozzle moving speed is dynamically planned and adjusted. Based on the spatial gradient characteristics of the thickness error distribution matrix, a fuzzy control strategy is used to generate the nozzle moving speed adjustment value. Based on the nonlinear coupling relationship between paint flow rate and spraying pressure, a compensation function model is constructed, and the paint flow rate compensation coefficient is calculated according to the spraying pressure adjustment value and the nozzle moving speed adjustment value. The spraying pressure adjustment value, the nozzle movement speed adjustment value, and the paint flow compensation coefficient are fused together to generate a comprehensive spraying parameter correction value.

9. The method for applying an antistatic coating to an LED circuit board according to claim 8, characterized in that, In step S4, the local coating thickness prediction matrix is ​​compared point by point with the target thickness distribution, and a global thickness deviation index is obtained based on the weighted integral algorithm and time decay processing. The spraying pressure adjustment value is then generated based on the global thickness deviation index using a PID adjustment algorithm.

10. The method for applying an antistatic coating to an LED circuit board according to claim 1, characterized in that, Step S5 specifically includes: Based on the calculated spraying pressure adjustment value, the opening of the pneumatic regulating valve is adjusted by PID to obtain an output pressure value that matches the target pressure. Based on the nozzle movement speed adjustment value, the drive frequency of the servo motor is adjusted using closed-loop feedback to obtain a nozzle scanning speed that matches the current working conditions. Based on the paint flow compensation coefficient, the output pulse width of the metering pump is proportionally adjusted. Multivariate coupling verification is performed on the adjusted spraying pressure, nozzle movement speed and paint flow rate to generate a comprehensive control state evaluation value; The adjusted spraying parameters are fed back to the fluid dynamics prediction model to update the model input variables, forming a closed-loop control data feedback mechanism.

Images