PCB three-proofing coating quality detection method and system

Abstract

The invention discloses a PCB three-proofing coating quality detection method and system. The method comprises the steps that a visible light reflection image of the surface of a PCB and scattering spectrum data of a near-infrared band are acquired; performing spatial filtering processing on the visible light reflection image to extract interface area pixels, calculating coating layer thickness gradient distribution based on Mie scattering characteristics in scattering spectrum data, and fusing to generate an edge gradient distribution map; and according to the gradient amplitude change rate in the edge gradient distribution map, segmenting the effective coverage area of the coating layer by adopting a self-adaptive dynamic threshold algorithm, and extracting curvature extreme point density and boundary fractal dimension parameters at the segmentation boundary. According to the invention, three technical barriers of difficult microdefect capture, unmeasurable interface performance and delayed failure risk in traditional detection are solved, and two-dimensional accurate diagnosis of the structural compactness and interface reliability of the coating layer is realized.

Description

Technical Field

[0001] This application relates to the technical field of PCB board detection, and particularly to a method and system for detecting the quality of three-proof coating on a PCB board. Background Art

[0002] With the development of miniaturization and high-density integration of electronic devices, the quality of the three-proof paint coating on a printed circuit board (PCB) directly affects the long-term reliability of the product in harsh environments such as high humidity and salt spray.

[0003] The existing Chinese patent with the publication number CN117994243B relates to the technical field of PCB board detection. Specifically, it is a method for detecting the coating quality of a PCB board. The collected PCB image is segmented to obtain a three-proof paint coverage area map; edge filtering processing is performed to obtain an image containing bubble edge information; bubble features are extracted to obtain a set of bubble feature maps; a bubble result set is calculated; and the bubble result set is classified to obtain conclusions and suggestions. Through the detection method of the PCB board coating quality, by segmenting the collected PCB image to obtain a three-proof paint coverage area map, and then performing edge filtering processing, extracting bubble features, calculating the bubble size and other steps to obtain the bubble situation of the PCB board, and classifying the bubble result set to obtain conclusions and suggestions, the detection of coating bubbles on the PCB board is realized, and the problem that the appearance of bubbles or blisters on the PCB affects the flatness and insulation performance of the coating is solved.

[0004] However, in the process of implementing the related technical solutions, it is found that there are at least the following technical problems: The above technology mainly relies on visible light imaging technology. The coating area is extracted through gray threshold segmentation, and macroscopic bubbles are detected by combining edge filtering. Although such methods are applied in the rapid detection of production lines, the detection system based on two-dimensional image morphology analysis can only identify macroscopic bubbles with a diameter of more than 50 μm, and lacks the detection ability for micron-level pores (<20 μm) inside the coating and weak interface bonding areas. And such micro-defects are the potential causes of failures such as dielectric breakdown and ion migration; in addition, the existing technology cannot quantitatively evaluate the interface bonding strength between the coating layer and the PCB substrate, resulting in the inability to warn of interface peeling failures caused by thermal stress cycling or mechanical vibration. Such hidden defects often appear after the equipment has been in service for several months, resulting in high after-sales maintenance costs. Summary of the Invention

[0005] In order to solve the above problems, an embodiment of the present invention provides a method for detecting the quality of three-proof coating on a PCB board, and the method includes: Obtain the visible light reflection image on the surface of the PCB and the scattering spectrum data in the near-infrared band; Perform spatial domain filtering on the visible light reflection image to extract the pixels in the interface region. At the same time, calculate the thickness gradient distribution of the coating layer based on the Mie scattering characteristics in the scattering spectral data, and fuse them to generate an edge gradient distribution map; According to the rate of change of the gradient amplitude in the edge gradient distribution map, use the adaptive dynamic threshold algorithm to segment the effective coverage area of the coating layer, and extract the density of the curvature extreme points and the boundary fractal dimension parameters at the segmentation boundary; Input the density of the curvature extreme points and the boundary fractal dimension parameters into the pre-trained coating curing morphology prediction model, and output the predicted value of the interface bonding strength of the coating layer and the coordinates of the potential peeling risk area by coupling the thermodynamic simulation data; Scan the surface of the coating layer with an array dielectric coupling sensor, measure the impedance phase response of each scanning point at different excitation frequencies, and calculate the microscopic porosity distribution map according to the slope of the phase offset varying with frequency; Perform spatio-temporal registration on the predicted value of the interface bonding strength, the coordinates of the potential peeling risk area, and the microscopic porosity distribution map, and generate a comprehensive evaluation matrix of the coating layer quality through a multi-source data decision fusion algorithm, and output the defect location report and the quality grade classification result.

[0006] Further, the spatial domain filtering method includes: Use the non-uniform illumination compensation algorithm to perform illumination equalization processing on the visible light reflection image. The algorithm calculates the pixel compensation coefficient by establishing a local illumination distribution model, where the compensation coefficient is dynamically adjusted according to the surface roughness characteristics in the near-infrared band scattering spectral data.

[0007] Further, the method for generating the edge gradient distribution map includes: Superimpose the gradient phase information of the visible light reflection image and the Mie scattering phase delay of the near-infrared scattering spectrum, and eliminate the interference of the substrate material texture on the coating boundary recognition through the phase gradient consistency verification.

[0008] Further, the adaptive dynamic threshold algorithm includes: Establish a probability density function according to the statistical distribution of the rate of change of the gradient amplitude, use the variational mode decomposition technique to separate the threshold response characteristics of the coating body and the boundary transition region, and dynamically determine the segmentation threshold interval.

[0009] Further, the training method of the coating curing morphology prediction model includes: Construct a finite element simulation model of multi-physics field coupling of heat, humidity and force. By introducing the time-varying viscoelastic constitutive equation of the coating material, generate a training set of the interface stress distribution under different curing conditions. The training set includes the mapping relationship between the temperature gradient, the humidity penetration depth and the interface peeling stress.

[0010] Furthermore, the analysis method of the impedance phase response includes: An equivalent circuit model of the microscopic pores in the coating layer was established, and the phase shift at different frequencies was fitted by the Cole-Cole distribution function to extract the pore connectivity index and the equivalent dielectric relaxation time parameters.

[0011] Furthermore, the implementation method of spatiotemporal registration includes: A three-dimensional registration coordinate system is established based on the spatial topological structure of the coating surface. The mechanical confidence of the predicted value of the interface bonding strength and the electrical sensitivity of the micro-porosity distribution are weightedly fused to generate a quality evaluation weight matrix with physical interpretability.

[0012] Furthermore, a method for detecting the quality of the conformal coating of a PCB board further includes: A multi-frequency eddy current sensing unit is integrated into the array dielectric coupling sensor, and the coating dielectric phase response and substrate eddy current skin effect signals are synchronously collected through frequency division multiplexing technology; Based on the inverse square relationship between the skin depth of eddy current signals and the frequency, a frequency domain blind source separation model is constructed to decouple the cross-coupled noise components caused by metal substrate defects from the dielectric phase response. The frequency domain characteristics of high-frequency eddy currents, which are sensitive to the pore closure of the coating surface, and low-frequency eddy currents, which characterize the continuity of the base metal, are used to reconstruct the microscopic porosity distribution map after removing the base artifacts.

[0013] A PCB board conformal coating quality inspection system, the system comprising: Dual-mode acquisition module, which acquires the visible light reflection image and the scattering spectrum data in the near-infrared band of the PCB surface; A gradient map generation module, which performs spatial domain filtering on the visible light reflection image to extract interface area pixels, and calculates the coating thickness gradient distribution based on the Mie scattering characteristics in the scattering spectrum data, and fuses them to generate an edge gradient distribution map; A dynamic threshold segmentation module, which uses an adaptive dynamic threshold algorithm to segment the effective coverage area of ​​the coating layer according to the gradient amplitude change rate in the edge gradient distribution map, and extracts the curvature extreme point density and boundary fractal dimension parameters at the segmentation boundary; An interface strength prediction module, which inputs the curvature extreme point density and boundary fractal dimension parameters into a pre-trained coating solidification morphology prediction model, and outputs a coating layer interface bonding strength prediction value and potential peeling risk area coordinates by coupling thermodynamic simulation data; The dielectric pore analysis module scans the surface of the coating layer through an array of dielectric coupling sensors, measures the impedance phase response of each scanning point at different excitation frequencies, and calculates the microscopic porosity distribution map according to the slope of the phase offset varying with frequency; The result output module performs spatio-temporal registration on the predicted value of the interface bonding strength, the coordinates of the potential peeling risk area, and the microscopic porosity distribution map, generates a comprehensive evaluation matrix of the coating layer quality through a multi-source data decision fusion algorithm, and outputs a defect location report and a quality grade classification result.

[0014] The technical effects and advantages of a three-proof coating quality detection method and system for PCB boards provided by the present invention: The present invention constructs a multi-modal sensing and dielectric-mechanical coupling analysis system, overcomes the three major technical barriers of "difficult to capture microscopic defects, impossible to measure interface performance, and late manifestation of failure risks" in traditional detection, and realizes two-dimensional accurate diagnosis of the coating layer structure compactness and interface reliability. The present invention jointly models the polarization phase topology mapping of near-infrared Mie scattering and the Cole-Cole distribution of dielectric relaxation spectra to establish a three-dimensional dielectric anisotropy characterization of the internal microstructure of the coating, breaks through the shielding effect of the optical diffraction limit on microscopic defects, realizes non-destructive visualization of the nano-scale heterogeneous interface in the coating-substrate transition region, and accurately locks the preferential path of ion migration; constructs a dynamic evolution model of the adhesion work of the coating-substrate interface based on the thermodynamic potential function, fuses the dispersion characteristics of surface acoustic waves and the entropy change correlation of the dielectric relaxation time spectrum, reveals the energy threshold for the initiation and propagation of interface cracks in thermal stress cycles, and realizes the identification of the subcritical state of interface peeling through the phase change characteristics of acoustic-electric dual-mode signals; uses the curvature manifold learning algorithm to deconstruct the surface fluctuation modes of the coating layer, and distinguishes real defects from substrate features through the phase gradient consistency verification of topological invariants, eliminates the optical aliasing effect caused by the roughness of the copper foil, and improves the recognition specificity of microcracks and pores. Description of the Drawings

[0015] Figure 1 It is a flowchart of a three-proof coating quality detection method for PCB boards in Embodiment 1; Figure 2 It is a flowchart of a three-proof coating quality detection method for PCB boards in Embodiment 2; Figure 3 It is a schematic connection diagram of a three-proof coating quality detection system for PCB boards in Embodiment 3. Detailed Embodiments

[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] Example 1: Please refer to Figure 1 As shown, an embodiment of the present invention provides a method for detecting the quality of a three-conformal coating of a PCB board, the method comprising: Obtain visible light reflection images and near-infrared scattering spectrum data of the PCB surface.

[0018] The visible light reflection image is subjected to spatial domain filtering to extract interface area pixels, and the coating layer thickness gradient distribution is calculated based on the Mie scattering characteristics in the scattering spectrum data, and the edge gradient distribution map is generated by fusion.

[0019] According to the gradient amplitude change rate in the edge gradient distribution map, an adaptive dynamic threshold algorithm is used to segment the effective coverage area of ​​the coating layer, and the curvature extreme point density and boundary fractal dimension parameters at the segmentation boundary are extracted.

[0020] The curvature extreme point density and boundary fractal dimension parameters are input into a pre-trained coating curing morphology prediction model, and the coating layer interface bonding strength prediction value and the coordinates of the potential peeling risk area are output by coupling thermodynamic simulation data.

[0021] The surface of the coating layer is scanned by an array dielectric coupling sensor, the impedance phase response of each scanning point at different excitation frequencies is measured, and the microscopic porosity distribution map is calculated based on the slope of the phase shift varying with frequency.

[0022] The predicted value of interface bonding strength, coordinates of potential peeling risk areas and micro porosity distribution map are temporally and spatially registered, and a comprehensive evaluation matrix of coating quality is generated through a multi-source data decision fusion algorithm, and a defect location report and quality grade classification results are output.

[0023] Visible light reflectance images are acquired using a CCD or CMOS sensor camera, while scattering spectral data in the near-infrared band can be acquired using a near-infrared spectrometer.

[0024] Spatial domain filtering processing methods include: A non-uniform illumination compensation algorithm is used to perform illumination equalization processing on the visible light reflection image. The algorithm calculates the pixel compensation coefficient by establishing a local illumination distribution model, wherein the compensation coefficient is dynamically adjusted according to the surface roughness characteristics in the near-infrared band scattering spectrum data.

[0025] Specific methods include: The visible light reflection image is divided into multiple local regions. In each region, a two-dimensional Gaussian distribution model is established based on the pixel brightness statistical value to fit the illumination attenuation trend of the region and generate an initial illumination compensation coefficient. However, since the surface roughness difference of the PCB substrate will affect the distribution of the specular component of visible light reflection, it is difficult to accurately distinguish the real coating boundary from the illumination artifacts directly based on the statistical model of the visible light image. Therefore, combined with the surface roughness characteristics extracted from the scattering spectral data in the near-infrared band, the compensation coefficient is dynamically corrected. The correction methods include: When the Mie scattering phase delay of the near-infrared scattering spectrum increases, it indicates that the surface roughness of the current region is relatively high. At this time, the spatial smoothing constraint weight of the compensation coefficient is enhanced to suppress the brightness mutation caused by micro-structure scattering in the high-roughness region. On the contrary, when the surface roughness is relatively low, the high-frequency texture details are preferentially retained to improve the recognition rate of the coating boundary.

[0026] The compensation coefficient after dynamic adjustment acts on the visible light reflection image, eliminating the influence of non-uniform illumination while retaining the real interface characteristics of the coating and the substrate, providing a more robust preprocessing result for the subsequent generation of the edge gradient distribution map.

[0027] The method for generating the edge gradient distribution map includes: The gradient phase information of the visible light reflection image is superimposed with the Mie scattering phase delay of the near-infrared scattering spectrum, and the interference of the substrate material texture on the coating boundary recognition is eliminated through the verification of phase gradient consistency.

[0028] In the process of generating the edge gradient distribution map, first, the gradient field of the visible light reflection image is calculated, and the phase direction angle information of each pixel point is extracted to characterize the coating edge trend; simultaneously, the Mie scattering phase delay in the near-infrared scattering spectral data is analyzed. The Mie scattering phase delay reflects the refractive index difference between the coating material and the substrate at the sub-wavelength scale. After spatially registering the two phase characteristics, a dual-channel phase superposition algorithm is used to generate a composite phase map. The logical basis of the dual-channel phase superposition algorithm includes: the visible light gradient phase information is used to identify the macroscopic edge trend, and its phase angle change rate is positively correlated with the coating thickness mutation region; the near-infrared Mie scattering phase delay enhances the phase jump characteristics at the coating-substrate interface by detecting the optical field interference effect at the dielectric interface.

[0029] When the substrate material has periodic textures (such as fiberglass woven structure), its visible light gradient phase will show regular fluctuations, but the corresponding Mie scattering phase delay remains stable because the substrate surface is not covered with a coating. Using this characteristic, the phase gradient consistency discrimination conditions are set, including: If the change rate of the visible light phase angle in a certain area exceeds a preset threshold (e.g., ±15°), while the change amplitude of the Mie scattering phase delay is lower than the critical value (e.g., <5°), then it is determined that the area is substrate texture interference and attenuation processing is performed in the composite phase diagram; when the fluctuation trends of the two phase characteristics are strongly correlated in space, it is confirmed as the real coating boundary and the phase superposition result is retained.

[0030] The composite phase diagram after being corrected by the phase gradient consistency discrimination is normalized and mapped to obtain the edge gradient distribution map.

[0031] The phase gradient consistency discrimination significantly suppresses the interference of the microstructure of the substrate material on the recognition of the coating boundary; for example, the pseudo - edge caused by the substrate texture may produce a phase angle mutation of 30° - 45° in the visible light phase diagram, but since the Mie scattering phase delay in the uncoated area only fluctuates by 2° - 3°, it will be effectively filtered out in the consistency verification stage.

[0032] The adaptive dynamic threshold algorithm includes: Establish a probability density function according to the statistical distribution of the change rate of the gradient amplitude, adopt the variational mode decomposition technology to separate the threshold response characteristics of the coating body and the boundary transition area, and dynamically determine the segmentation threshold interval.

[0033] The adaptive dynamic threshold algorithm specifically includes: By statistically analyzing the change rate of the gradient amplitude of all pixel points in the map, it can be found that its probability density function shows a bimodal distribution characteristic. The main peak corresponds to the low - gradient fluctuation in the coating body area, and the secondary peak represents the high - gradient transition in the boundary transition area. However, in actual detection, the two peaks often partially overlap due to the gradual change of the coating thickness or the interference of substrate residues, resulting in difficulty in accurately segmenting with a single threshold. Therefore, the variational mode decomposition technology is introduced to decompose the original gradient signal into multiple modal components, including the first modal component, the second modal component, and the residual component.

[0034] The first modal component captures the low - frequency slow - change characteristics of the coating body area, and its amplitude envelope is related to the material uniformity.

[0035] The second modal component extracts the intermediate - frequency oscillation characteristics of the boundary transition area, reflecting the interface mutation between the coating layer and the substrate.

[0036] The residual component includes high - frequency interference introduced by noise and substrate micro - defects.

[0037] Calculate the energy proportion of each modal component to determine whether the current analysis area is near the effective boundary.

[0038] Taking the initial bimodal center value as a constraint condition, the KL divergence minimization criterion is adopted to adjust the frequency band width of the modal components, so that the decomposed modal components can accurately separate the coating body and boundary features in the time-frequency domain.

[0039] According to the processed modal components above, reconstruct the threshold response curve and dynamically determine the segmentation threshold interval. Exemplarily: When it is detected that there is a gradual change in thickness in a certain local area (manifested as the continuous increase in the amplitude of the second modal component), the algorithm adjusts the lower threshold from 0.35 of the standard value to 0.28 to expand the boundary capture range; conversely, for the high-uniformity coating area (the proportion of the first modal energy > 70%), the upper threshold is tightened to suppress the generation of pseudo boundaries.

[0040] The training method of the coating curing morphology prediction model includes: Construct a finite element simulation model of multi-physical field coupling (heat-moisture-force). By introducing the time-varying viscoelastic constitutive equation of the coating material, generate a training set of interface stress distributions under different curing conditions. The training set includes the mapping relationship between the temperature gradient, humidity penetration depth and interface peeling stress.

[0041] In the training stage of the coating curing morphology prediction model, construct a finite element simulation framework of heat-moisture-force coupling. The construction method of the finite element simulation framework includes: Thermal field modeling: Characterize the evolution of the temperature gradient inside the coating during the curing process through the unsteady heat conduction equation, and consider the non-linear effects of factors such as the heating rate and environmental heat dissipation on the temperature distribution.

[0042] Moisture field modeling: Simulate the moisture penetration process using Fick's diffusion law, and introduce a dynamic change model of the coating porosity to reflect the time-varying characteristics of the moisture absorption performance during the curing stage.

[0043] Mechanical modeling: Combining the time-varying viscoelastic constitutive equation, express the elastic modulus and relaxation time parameters of the coating as functions of the degree of curing and temperature-humidity conditions, and accurately describe the stress relaxation behavior of the material during the transformation from liquid to solid state.

[0044] Through the collaborative solution of the above models, generate a dataset of interface stress distributions covering different process conditions; Exemplarily: In the simulation, set a non-uniform distribution of the temperature gradient from 25°C to 80°C, and the humidity penetration depth changes in layers within the range of 0.1mm to 0.5mm. Output the component data of the corresponding peeling stress in the normal and tangential directions of the interface. This component data forms a structured training set through parametric sampling, where each data includes an input vector (temperature gradient extreme value, humidity penetration characteristic length and curing time) and an output vector (maximum peeling stress amplitude and critical peeling angle).

[0045] To address the problem of model generalization caused by the uncertainty of coating material parameters in actual detection, a material property perturbation mechanism is introduced in the simulation stage. The material property perturbation mechanism includes: Apply a ±15% random fluctuation to the key parameters such as activation energy and free volume fraction in the viscoelastic constitutive equation to generate virtual samples with statistical diversity.

[0046] During the training process, a bidirectional long short-term memory network is used to capture the temporal correlations of multi-physical field parameters, and an attention mechanism is designed to strengthen the cross-scale correlation feature extraction of humidity penetration and stress accumulation. The trained coating curing morphology prediction model can, based on the temperature and humidity distribution data obtained from on-line detection (such as infrared thermography and near-infrared humidity sensing results), predict the spatial distribution trend of interfacial peeling stress in real time.

[0047] The analysis method of impedance phase response includes: Establish an equivalent circuit model for the microscopic pores of the coating layer, and fit the phase offset at different frequencies through the Cole-Cole distribution function to extract the pore connectivity index and the equivalent dielectric relaxation time parameter.

[0048] The specific analysis method includes: First, construct a hierarchical equivalent circuit model according to the microscopic morphology characteristics after the coating is cured. The hierarchical equivalent circuit model includes a surface conduction path, a bulk dielectric layer, and an interface transmission channel.

[0049] Surface conduction path: A resistance network formed by the conductive carbonized products at the pore edges, and its resistance value is negatively correlated with the pore opening rate.

[0050] Bulk dielectric layer: The uncured coating body is equivalent to a parallel resistor-capacitor unit, and the capacitance component reflects the polarization characteristics of the matrix material.

[0051] Interface transmission channel: The ion migration path at the coating-substrate interface is modeled as a series of constant phase angle elements, which characterizes the non-ideal charge transfer behavior.

[0052] Obtain the phase offset data in the range of 10 Hz to 1 MHz through broadband impedance spectroscopy measurement, and perform a non-linear fitting on the frequency response curve using the Cole-Cole distribution function. The Cole-Cole distribution function can, by introducing the relaxation time distribution parameter, simultaneously characterize the multi-scale relaxation effects of the pore structure, including: High frequency band (>100 kHz): Corresponding to the dipole orientation polarization of nanoscale isolated pores, the equivalent dielectric relaxation time is obtained by fitting. When its value is less than 1 μs, it indicates that the pores are well closed.

[0053] Mid - frequency band (1 kHz - 100 kHz): It reflects the ion diffusion relaxation of micron - level connected pores. The closer the extracted pore connectivity index α is to 1, the higher the penetration of the pore network.

[0054] Low - frequency band (<1 kHz): It is related to the interfacial polarization effect of macroscopic interface defects and is used to correct the boundary conditions of the equivalent circuit model.

[0055] Exemplary: When analyzing a polyester coating sample, the fitting results show that the equivalent dielectric relaxation time = 0.85 μs and α = 0.76. This indicates that although there are some sub - micron - level pores in the coating (with a shorter equivalent dielectric relaxation time), the connectivity index is lower than the qualified threshold of 0.8, and a quality warning needs to be triggered. This result can be cross - verified with the interface stress prediction result. When α < 0.8, the probability that the corresponding interfacial peeling stress peak exceeds the critical value of the material adhesion strength increases to 92%.

[0056] The implementation methods of spatio - temporal registration include: Based on the surface space topological structure of the coating layer, a three - dimensional registration coordinate system is established. The mechanical confidence of the predicted interfacial bonding strength value and the electrical sensitivity of the microscopic porosity distribution are weighted and fused to generate a physically interpretable quality evaluation weight matrix.

[0057] Specific implementation methods include: According to the coating surface curvature distribution (obtained by analyzing the density of curvature extreme points), a non - uniform meshed three - dimensional registration coordinate system is established, where the spatial resolution of each grid node is aligned with the pixel scale of multi - spectral imaging to achieve physical data association, including: Project the predicted interfacial bonding strength distribution data onto the three - dimensional coordinate system. Based on the spatial correlation analysis of the stress gradient, calculate the mechanical confidence parameters at each node to reflect the probability distribution of the interfacial peeling risk.

[0058] Normalize the pore connectivity index and the dielectric relaxation time parameter, and combine the spatial topological relationship of the conductive channels in the equivalent circuit model to generate a distribution heat map representing the electrical response sensitivity of microscopic defects.

[0059] Design a two - parameter adaptive weight function. The mechanical confidence weight increases non - linearly as the standard deviation of the predicted interfacial bonding strength value increases, and the electrical sensitivity weight is positively correlated with the local gradient norm of the porosity distribution. For example, in the region where the coating edge curvature changes suddenly (curvature radius < 0.3 mm), the mechanical weight is increased to 0.7 to give priority to responding to the interfacial peeling risk; while in the central uniform region of the coating (porosity > 5%), the electrical weight is increased to 0.6 to strengthen the detection of microscopic defects.

[0060] Spatial continuity optimization of electrical sensitivity weights and mechanical confidence weights is achieved through Gaussian process regression to generate a physically interpretable quality evaluation weight matrix. Each unit value of the quality evaluation weight matrix represents the comprehensive quality coefficient of the corresponding spatial position. When the comprehensive quality coefficient is lower than 0.4 (example threshold), hierarchical early warning is triggered. The hierarchical early warning criteria include: When the comprehensive quality coefficient is in the range of 0.3 - 0.4, it indicates the need for fine-tuning of local process parameters.

[0061] When the comprehensive quality coefficient is less than 0.3, it is determined as a high-risk area of structural failure, and adaptive detection needs to be initiated for re-inspection.

[0062] The multi-source data decision fusion algorithm includes: Using the evidence reasoning theory to construct a D-S confidence assignment framework, converting the boundary integrity score into a basic probability assignment function, and performing uncertainty reasoning in combination with the prior distribution function of the interface bonding force level to output the confidence interval evaluation result of defect location.

[0063] The multi-source data decision fusion algorithm specifically includes: Establish an identification framework for coating quality evaluation, define three basic propositions of "interface bonding failure", "pore densification defect" and "composite damage", and realize decision-making reasoning through the following steps: Map the boundary integrity score (curvature extreme point density and gradient direction consistency parameter) to the initial belief assignment.

[0064] When the high-density curvature area (>15 points / mm²) and the direction dispersion >60°, assign a basic probability of 0.6 to the proposition of "interface bonding failure".

[0065] Low pore connectivity index (α < 0.7) accompanied by short dielectric relaxation time , assign a belief of 0.55 to the proposition of "pore densification defect".

[0066] The remaining uncertain part is retained as the global conflict factor for conflict resolution during subsequent evidence synthesis.

[0067] Combined with the interface bonding force level distribution, establish a prior probability correction model based on the Beta function; for example, when the standard deviation of the adhesion strength of the batch material exceeds 0.8 MPa, increase the prior weight of the proposition of "composite damage" to 0.3 to reflect the influence of the inherent performance fluctuation of the material on the defect mode.

[0068] For areas with high conflict (global conflict factor K > 0.4), trigger an adaptive weighted correction mechanism, and the adaptive weighted correction mechanism includes: Introduce the spatial confidence parameter in the quality evaluation weight matrix to perform local weighted average processing on the conflicting evidence.

[0069] The final output is a three-dimensional confidence interval heat map, in which each voxel contains a confidence distribution in the interval [0.2, 0.85]. When the confidence of "interface bonding failure" exceeds 0.7 and the spatial continuous area is >4mm², it is judged as a high-risk defect cluster.

[0070] Embodiment 2: Figure 2 As shown, this embodiment further improves the design on the basis of embodiment 1. The difference is that in actual operation, it is found that in embodiment 1, only the dielectric coupling sensor is relied on to detect the porosity, but the electromagnetic field distortion caused by the inherent defects of the metal substrate (such as microcracks and grain boundary segregation) is not considered, resulting in the aliasing of the pore signal and the substrate artifact. Based on this, a PCB board three-proof coating quality detection method also includes: A multi-frequency eddy current sensing unit is integrated in the array dielectric coupling sensor, and the coating dielectric phase response and substrate eddy current skin effect signals are synchronously collected through frequency division multiplexing technology.

[0071] A frequency domain blind source separation model is constructed based on the inverse square relationship between the skin depth of eddy current signals and the frequency, and the cross-coupling noise components caused by metal substrate defects are decoupled from the dielectric phase response.

[0072] The frequency domain characteristics of high-frequency eddy currents, which are sensitive to the pore closure of the coating surface, and low-frequency eddy currents, which characterize the continuity of the base metal, are used to reconstruct the microscopic porosity distribution map after removing the base artifacts.

[0073] Furthermore, the multi-frequency eddy current sensing unit includes a dielectric sensor and an eddy current sensing unit; the dielectric sensor measures the phase shift response of the dielectric constant of the coating layer by applying a high-frequency alternating electric field (typical excitation frequency is 1MHz-10MHz), while the eddy current sensing unit adopts a wide-band scanning mode (typical coverage range is 50kHz-5MHz) to obtain the skin effect signal of the base metal through the principle of electromagnetic induction; the two types of sensors adopt a coplanar electrode design to ensure the temporal and spatial synchronization of the dielectric-eddy current signals and avoid the registration error introduced by the scanning position offset.

[0074] Furthermore, by extracting the impedance mutation characteristics reflecting the grain boundary cracks or oxide layers of the substrate metal from the low-frequency eddy current signal (such as below 100kHz), the artifact noise template of the substrate defect is reconstructed; at the same time, the high sensitivity of the high-frequency eddy current signal (such as above 1MHz) to the near-surface pore closure of the coating is utilized to establish the frequency domain orthogonality criterion of the pore-substrate signal; the cross-coupling components related to the substrate artifacts are separated from the mixed dielectric phase spectrum through the independent component analysis (ICA) algorithm to achieve dynamic noise suppression of the real pore response of the coating.

[0075] Furthermore, the porosity distribution reconstruction is optimized according to the physical property differences of the multi-frequency eddy current signals. The reconstruction method includes: Due to the relatively shallow skin depth, high-frequency eddy currents mainly capture the closed state of micron-sized pores on the coating surface, and the signal amplitude is negatively correlated with the pore gas filling rate; low-frequency eddy currents penetrate the coating and reach the substrate, excluding the interference of substrate cracks on the coating structure evaluation through metal continuity detection; the denoised dielectric phase slope and the high-frequency eddy current amplitude are weighted and fused to generate a microscopic porosity distribution map resistant to substrate interference, forming a spatial correlation mapping with the interface bonding strength, and jointly input into the multi-source decision fusion engine for comprehensive quality evaluation.

[0076] Example 3: As Figure 3 shown, based on the same inventive concept as a method for detecting the quality of three-proof coating on a PCB board in the foregoing embodiments, the present application provides a system for detecting the quality of three-proof coating on a PCB board. The system in the embodiments of the present application and the method embodiments are based on the same inventive concept. Among them, the system includes: A dual-modal acquisition module that acquires the visible light reflection image and the scattering spectral data in the near-infrared band on the surface of the PCB.

[0077] A gradient map generation module that performs spatial domain filtering on the visible light reflection image to extract interface region pixels, and at the same time calculates the thickness gradient distribution of the coating layer based on the Mie scattering characteristics in the scattering spectral data, and fuses them to generate an edge gradient distribution map.

[0078] A dynamic threshold segmentation module that, according to the gradient amplitude change rate in the edge gradient distribution map, uses an adaptive dynamic threshold algorithm to segment the effective coverage area of the coating layer, and extracts the density of curvature extreme points and the boundary fractal dimension parameters at the segmentation boundary.

[0079] An interface strength prediction module that inputs the density of curvature extreme points and the boundary fractal dimension parameters into a pre-trained coating curing morphology prediction model, and outputs the predicted value of the coating interface bonding strength and the coordinates of the potential peeling risk area through coupling thermodynamic simulation data.

[0080] A dielectric pore analysis module that scans the surface of the coating layer through an array of dielectric coupling sensors, measures the impedance phase response of each scan point at different excitation frequencies, and calculates the microscopic porosity distribution map according to the slope of the phase offset change with frequency.

[0081] A result output module that spatially and temporally registers the predicted value of the interface bonding strength, the coordinates of the potential peeling risk area, and the microscopic porosity distribution map, generates a comprehensive evaluation matrix of the coating layer quality through a multi-source data decision fusion algorithm, and outputs a defect location report and a quality grade classification result.

[0082] Obviously, those skilled in the art can make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

[0083] The above are only the preferred specific embodiments of the embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed in the present application, according to the technical solution and its concept of the present application, makes equivalent substitutions or changes, and should be covered by the protection scope of the present application.

Claims

1. A method for detecting the quality of a PCB conformal coating, characterized in that: include: Obtain visible light reflection images and near-infrared scattering spectrum data of the PCB surface; Performing spatial domain filtering on the visible light reflection image to extract pixels in the interface area, and calculating the coating layer thickness gradient distribution based on the Mie scattering characteristics in the scattering spectrum data, and fusing them to generate an edge gradient distribution map; According to the gradient amplitude change rate in the edge gradient distribution map, an adaptive dynamic threshold algorithm is used to segment the effective coverage area of ​​the coating layer, and the curvature extreme point density and boundary fractal dimension parameters at the segmentation boundary are extracted; Inputting the curvature extreme point density and boundary fractal dimension parameters into a pre-trained coating curing morphology prediction model, and outputting a coating layer interface bonding strength prediction value and potential peeling risk area coordinates by coupling thermodynamic simulation data; The coating surface is scanned by an array dielectric coupling sensor, the impedance phase response of each scanning point at different excitation frequencies is measured, and the microscopic porosity distribution map is calculated according to the slope of the phase shift varying with frequency; The predicted value of interface bonding strength, coordinates of potential peeling risk areas and micro porosity distribution map are temporally and spatially registered, and a comprehensive evaluation matrix of coating quality is generated through a multi-source data decision fusion algorithm, and a defect location report and quality grade classification results are output.

2. A PCB board three-conformal coating quality detection method according to claim 1, characterized in that: Spatial domain filtering processing methods include: A non-uniform illumination compensation algorithm is used to perform illumination equalization processing on the visible light reflection image. The algorithm calculates the pixel compensation coefficient by establishing a local illumination distribution model, wherein the compensation coefficient is dynamically adjusted according to the surface roughness characteristics in the near-infrared band scattering spectrum data.

3. A PCB board three-conformal coating quality detection method according to claim 1, characterized in that: The method for generating the edge gradient distribution map includes: The gradient phase information of the visible light reflection image is superimposed with the Mie scattering phase delay of the near-infrared scattering spectrum, and the interference of the substrate material texture on the coating boundary identification is eliminated through phase gradient consistency verification.

4. A PCB board three-conformal coating quality detection method according to claim 1, characterized in that: The adaptive dynamic threshold algorithm includes: A probability density function is established according to the statistical distribution of the gradient amplitude change rate. The variational mode decomposition technique is used to separate the threshold response characteristics of the coating body and the boundary transition zone, and the segmentation threshold interval is dynamically determined.

5. A PCB board three-conformal coating quality detection method according to claim 1, characterized in that: The training method of the coating curing morphology prediction model includes: A finite element simulation model of heat-humidity-force multi-physics field coupling is constructed, and a training set of interface stress distribution under different curing conditions is generated by introducing the time-varying viscoelastic constitutive equation of the coating material. The training set includes the mapping relationship between temperature gradient, moisture penetration depth and interface peeling stress.

6. A PCB board three-conformal coating quality detection method according to claim 1, characterized in that: The analysis methods of impedance phase response include: An equivalent circuit model of the microscopic pores in the coating layer was established, and the phase shift at different frequencies was fitted by the Cole-Cole distribution function to extract the pore connectivity index and equivalent dielectric relaxation time parameters.

7. A PCB board three-conformal coating quality detection method according to claim 1, characterized in that: The implementation methods of spatiotemporal registration include: A three-dimensional registration coordinate system is established based on the spatial topological structure of the coating surface. The mechanical confidence of the predicted value of the interface bonding strength and the electrical sensitivity of the micro-porosity distribution are weightedly fused to generate a quality evaluation weight matrix with physical interpretability.

8. A PCB board conformal coating quality detection method according to claim 1, characterized in that: Also includes: A multi-frequency eddy current sensing unit is integrated into the array dielectric coupling sensor, and the coating dielectric phase response and substrate eddy current skin effect signals are synchronously collected through frequency division multiplexing technology; A frequency domain blind source separation model is constructed based on the inverse square relationship between the skin depth of eddy current signals and the frequency, and the cross-coupling noise components caused by metal substrate defects are decoupled from the dielectric phase response. The frequency domain characteristics of high-frequency eddy currents, which are sensitive to the pore closure of the coating surface, and low-frequency eddy currents, which characterize the continuity of the base metal, are used to reconstruct the microscopic porosity distribution map after removing the base artifacts.

9. A PCB board three-conformal coating quality inspection system, characterized in that: The system includes: Dual-mode acquisition module, which acquires the visible light reflection image of the PCB surface and the scattering spectrum data in the near-infrared band; A gradient map generation module, which performs spatial domain filtering on the visible light reflection image to extract interface area pixels, and calculates the coating thickness gradient distribution based on the Mie scattering characteristics in the scattering spectrum data, and fuses them to generate an edge gradient distribution map; A dynamic threshold segmentation module, which uses an adaptive dynamic threshold algorithm to segment the effective coverage area of ​​the coating layer according to the gradient amplitude change rate in the edge gradient distribution map, and extracts the curvature extreme point density and boundary fractal dimension parameters at the segmentation boundary; An interface strength prediction module, which inputs the curvature extreme point density and boundary fractal dimension parameters into a pre-trained coating solidification morphology prediction model, and outputs a coating layer interface bonding strength prediction value and potential peeling risk area coordinates by coupling thermodynamic simulation data; The dielectric pore analysis module scans the surface of the coating layer through an array dielectric coupling sensor, measures the impedance phase response of each scanning point at different excitation frequencies, and calculates the microscopic porosity distribution map according to the slope of the phase shift varying with frequency; The result output module performs spatiotemporal registration on the interface bonding strength prediction value, the coordinates of the potential peeling risk area and the micro porosity distribution map, generates a comprehensive evaluation matrix of the coating quality through a multi-source data decision fusion algorithm, and outputs a defect location report and a quality grade classification result.

Images