Circuit board deep hole surface uniformity control method

Abstract

The invention discloses a circuit board deep hole surface uniformity control method, and relates to the technical field of circuit board quality control, and the method comprises the following steps: a laser scanning system carries out the defect detection of the deep hole surface of a to-be-detected circuit board at a preset sensitivity, rapidly screens out a defect region, and guarantees the overall detection efficiency at the same time. According to the invention, through dynamic adjustment of sensitivity and deep learning intelligent evaluation, the detection precision of tiny defects is improved, and the problem of missing detection of a traditional fixed sensitivity system is solved. The sensitivity is evaluated and adjusted in real time, it is ensured that defects such as tiny cracks and bubbles can be captured, the omission ratio is reduced, and the stability and reliability of a high-frequency and high-performance circuit board are ensured. Meanwhile, the initial sensitivity setting ensures that obvious defects are rapidly screened, and production delay is avoided; for fine defects, the sensitivity is dynamically improved by the system, the defects are accurately identified, excessive detection is avoided, the production efficiency and quality control are balanced, and the efficiency of the whole production line and the product quality are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of printed circuit board quality control, and particularly to a method for controlling the surface uniformity of deep holes in a printed circuit board. Background Art

[0002] The control of the surface uniformity of deep holes in a printed circuit board refers to analyzing and optimizing the physical properties of the inner wall of the deep hole before the electroplating process to ensure the smoothness of the hole wall surface and the consistency of the thickness. The key to this process lies in identifying and eliminating defects caused by drilling or pretreatment processes, such as burrs, rough areas, or uneven coating on the hole wall. These problems may lead to uneven distribution of the electroplating layer thickness, thereby affecting the electrical conductivity, mechanical strength of the printed circuit board, and the reliability of subsequent processes. Therefore, the control of the surface uniformity of deep holes is an important precondition for ensuring high-quality standards in the deep hole area, and its core goal is to provide a stable and uniform substrate surface for electroplating.

[0003] The prior art has the following deficiencies:

[0004] In the prior art, a laser scanning system usually captures changes in the surface quality of deep holes in a printed circuit board by presetting the sensitivity, so as to optimize the production efficiency while ensuring the detection accuracy. However, the setting of a fixed sensitivity has problems of insufficient adaptability when facing micro-cracks, air bubbles, or other small defects. In this case, the laser scanning system may not be able to effectively identify these subtle defects because the change amplitude of the reflection signal of micro-cracks, air bubbles, or slight surface unevenness is usually lower than the threshold of the fixed sensitivity, resulting in the defects being missed. This not only affects the quality of the final product but also may lead to long-term potential risks. The missed defects may gradually expand with the change of the usage time and the external environment (such as temperature, humidity), and finally cause serious electrical failures or malfunctions. Especially in the application of high-frequency and high-performance circuit boards, missed detection may lead to problems such as short circuits and signal interference, seriously affecting the stability and reliability of the device.

[0005] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present disclosure, and therefore it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0006] The object of the present invention is to provide a method for controlling the surface uniformity of deep holes in a circuit board. By dynamically adjusting the sensitivity and intelligent evaluation through deep learning, the detection accuracy of micro-defects is improved, and the problem of missed detection in traditional fixed-sensitivity systems is solved. The sensitivity is evaluated and adjusted in real time to ensure that defects such as fine cracks and bubbles can be captured, reducing the missed detection rate and ensuring the stability and reliability of high-frequency and high-performance circuit boards. At the same time, the preliminary sensitivity setting ensures the rapid screening of obvious defects to avoid production delays; for micro-defects, the system dynamically increases the sensitivity to accurately identify the defects and avoid over-detection, balancing production efficiency and quality control, and improving the efficiency of the overall production line and the product quality, so as to solve the problems in the above-mentioned background technology.

[0007] To achieve the above object, the present invention provides the following technical solution: A method for controlling the surface uniformity of deep holes in a circuit board, comprising the following steps:

[0008] The laser scanning system performs defect detection on the surface of the deep holes of the circuit board to be detected with a preset sensitivity, quickly screening out the defective areas while ensuring the overall detection efficiency;

[0009] During the defect detection process, the laser scanning system real-time obtains the defect data information on the surface of the deep holes, providing a rich data basis for subsequent preprocessing and feature extraction;

[0010] Preprocess the obtained defect data information to improve the data quality and the accuracy of analysis;

[0011] In the preprocessed defect information, extract the key features reflecting the micro-defects on the surface of the current circuit board deep holes, and conduct a detailed analysis of the extracted key features under the detection window to quantify the defect degree of the surface quality of the current circuit board deep holes;

[0012] Input the analyzed key features into a pre-trained deep learning model, and conduct an intelligent evaluation of the quality defects of the current circuit board deep holes through the deep learning model;

[0013] According to the evaluation results of the deep learning model, divide the quality defects of the current circuit board deep holes into two categories: "micro-defects" and "significant defects";

[0014] For significant defects, continue to perform quality detection with the preset sensitivity to ensure rapid detection and correction and avoid production delays;

[0015] For micro-defects, based on the evaluation results of the deep learning model, dynamically increase the detection sensitivity to re-identify the current deep hole defects with a higher sensitivity.

[0016] Preferably, in the preprocessed defect information, key features reflecting the subtle defects on the surface of the current circuit board deep holes are extracted. The extracted features include the change in the reflection signal in the defect edge region and the change in the phase of the reflection signal. Under the detection window, the change in the reflection signal in the defect edge region and the change in the phase of the reflection signal are analyzed to generate a defect edge transition reference value and a phase change reference value respectively. The degree of defect on the surface quality of the current circuit board deep holes is quantified through the defect edge transition reference value and the phase change reference value.

[0017] Preferably, the specific steps for analyzing the change in the reflection signal in the defect edge region under the detection window to generate a defect edge transition reference value are as follows:

[0018] The surface of the circuit board deep holes is scanned by a laser scanning system to obtain the reflection signal, and the region containing the defect is determined. The calculation expression is as follows:

[0019]

[0020] , where R edge is the reflection signal intensity in the defect edge region, I(z, θ) is the reflection intensity signal, indicating the signal intensity of the laser beam reflected back at the position z on the surface of the deep hole and the scanning angle θ, Z 1 is the starting position of the defect region, Z 2 is the ending position of the defect region, θ 1 is the starting angle of the laser scan, θ 2 is the ending angle of the laser scan;

[0021] After determining the defect edge region, analyze the change trend of the reflection signal in this region, and quantify the change amplitude by calculating the derivative of the signal change. The calculation expression is as follows:

[0022]

[0023] , where Slope edge is the slope of the change in the reflection signal in the defect edge region, I(z 2 , θ 2 ) is the reflection signal intensity at the end position (z 2 , θ 2 ) of the defect region, I(z 1 , θ 1 ) is the reflection signal intensity at the starting position (z 1 , θ 1 ) of the defect region;

[0024] In the edge region of the deep hole surface defect, the change in the reflection signal is not linear. To accurately reflect the subtle degree of the defect, a non - linear mapping function is introduced to quantify the signal transition. The calculation expression is as follows:

[0025]

[0026] , where Transition exp is the defect edge transition intensity, k i is the weighting coefficient at the i-th position z i , I(z i , θ j ) refers to the reflected signal intensity at the i-th position z and the j-th angle θ, I 0 is the reference signal intensity, and n is the total number of positions;

[0027] The reflected signal intensity R edge passing through the defect edge region, the slope Slope edge of the change in the reflected signal within the defect edge region, and the defect edge transition intensity Transition exp are used to generate a defect edge transition reference value, and the generation formula is as follows:

[0028]

[0029] , where Defect Edge Transition is the defect edge transition reference value.

[0030] Preferably, the specific steps for analyzing the phase change of the reflected signal under the detection window to generate a phase change reference value are as follows:

[0031] For each scanning point of the laser scanning system, the phase data of the reflected signal is measured, and the formula for the phase data of the reflected signal is as follows:

[0032]

[0033] , where is the phase of the reflected signal at the coordinate point (x, y), A(x, y) is the amplitude distribution of the reflected signal at the coordinate point (x, y) on the surface of the deep hole of the circuit board, ω is the angular frequency of the laser beam, t is the time variable, j is the imaginary unit, and arg is the phase extraction function;

[0034] By performing a difference analysis on the phase data of each adjacent scanning point, the local phase change on the surface of the deep hole is calculated, which is achieved by calculating the phase difference between adjacent points, and the calculation expression is as follows:

[0035]

[0036] , where is the phase of the reflected signal at the coordinate point (x + 1, y + 1), referring to the phase of the point adjacent to the current point (x, y), is the phase difference between adjacent points, that is, the phase difference between the coordinate points (x, y) and the adjacent coordinate points (x + 1, y + 1), which is the difference in phase between the two points;

[0037] After calculating the phase difference between adjacent points the phase difference is quantified to obtain the local phase change degree. To quantify the phase change, a phase perturbation amplitude coefficient is introduced, and the calculation formula is as follows:

[0038]

[0039] , where C PD (x, y) is the local phase change degree, p is the adjustment index, is the maximum observed phase change amplitude;

[0040] All local phase change degrees C PD (x, y) are weighted and aggregated to generate a phase change reference value, and the generation formula is as follows:

[0041] Phase Shift =∫∫ Ω exp(-αC PD (x,y))dx dy

[0042] , where Ω is the detection area, α is the attenuation factor, and Phase Shift is the phase change reference value.

[0043] Preferably, after analyzing the extracted key features, the generated defect edge transition reference value and phase change reference value are input into a pre-trained deep learning model, and a defect type evaluation index is generated through the deep learning model. The current printed circuit board deep hole quality defect is intelligently evaluated through the defect type evaluation index.

[0044] Preferably, the defect type evaluation index generated when the current printed circuit board deep hole quality defect is intelligently evaluated through a pre-trained deep learning model is compared and analyzed with a pre-set defect type evaluation index reference threshold, and the current printed circuit board deep hole quality defect is classified. The classification steps are as follows:

[0045] If the defect type evaluation index is greater than or equal to the pre-set defect type evaluation index reference threshold, the current printed circuit board deep hole quality defect is classified as a minor defect;

[0046] If the defect type evaluation index is less than the pre-set defect type evaluation index reference threshold, the current printed circuit board deep hole quality defect is classified as a significant defect.

[0047] Preferably, for minor defects, based on the evaluation results of the deep learning model, the detection sensitivity is dynamically increased, and the specific steps to re-identify the current deep hole defects with higher sensitivity are as follows:

[0048] For the area determined to be a minor defect, the detection sensitivity is dynamically adjusted according to the current defect information. Based on the deep learning evaluation results, the sensitivity adjustment factor is calculated, and the calculation expression is as follows:

[0049]

[0050] , where SAF is the sensitivity adjustment factor, Defect Type ref is the reference threshold of the defect type evaluation index, Defect Type Assessment is the defect type evaluation index, μ is the factor controlling the exponential growth, γ is the decay factor, and ΔSignal is the change amplitude of the reflected signal on the deep hole surface;

[0051] Based on the sensitivity adjustment factor SAF, the sensitivity is reset and the quality of the current printed circuit board deep hole surface is detected. The calculation expression is as follows:

[0052]

[0053] , where S preset is the preset sensitivity, η is the weighting factor for sensitivity adjustment, δ is the flexible factor controlling sensitivity adjustment, ζ is the non-linear index of sensitivity adjustment, and S adjusted is the adjusted sensitivity.

[0054] In the above technical solution, the technical effects and advantages provided by the present invention are as follows:

[0055] The present invention significantly improves the detection accuracy of minor defects through dynamic sensitivity adjustment and intelligent evaluation based on deep learning. The traditional laser scanning system with fixed sensitivity cannot effectively identify minor defects, resulting in missed detection problems. However, this method can ensure that even microcracks, bubbles, or slight non-uniformities with extremely small change amplitudes of the reflected signal can be captured by real-time evaluation and dynamic sensitivity adjustment. By extracting key features reflecting minor defects (such as changes in the reflected signal and phase changes at the defect edge) and combining them with a deep learning model for precise evaluation, the system can significantly reduce the missed detection rate and ensure the reliability and stability of high-frequency and high-performance circuit boards.

[0056] While ensuring high-precision detection, the present invention effectively balances production efficiency and quality control. The preliminary sensitivity setting is used for rapid screening of obvious defects, avoiding production delays caused by over-detection. For subtle defects, the system dynamically increases the sensitivity according to the deep learning evaluation results, thus accurately identifying those defects missed by traditional methods while avoiding production process bottlenecks caused by over-detection. In this way, the production line can not only operate efficiently, but also ensure the quality and long-term reliability of the final product. Especially in high-demand application environments, it can effectively prevent potential electrical failures or signal interferences. Brief Description of the Drawings

[0057] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0058] Figure 1 It is a method flow chart of a method for controlling the surface uniformity of deep holes on a circuit board according to the present invention. Detailed Embodiments

[0059] Now, the exemplary embodiments will be described more comprehensively with reference to the accompanying drawings. However, the exemplary embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these exemplary embodiments are provided so that the present disclosure will be more complete and comprehensive, and the concept of the exemplary embodiments will be fully conveyed to those skilled in the art.

[0060] The present invention provides a method for controlling the surface uniformity of deep holes on a circuit board as shown in Figure 1 the following, including the following steps:

[0061] The laser scanning system performs defect detection on the surface of the deep holes of the circuit board to be detected with a preset sensitivity, quickly screening out the defective areas while ensuring the overall detection efficiency;

[0062] The preset sensitivity is set based on past experience and product quality requirements, and can capture larger defects, such as obvious cracks, bubbles or surface non-uniformities, without affecting the production speed. The setting of the preset sensitivity takes into account the noise level of the production environment and the reflection characteristics of the surface material to ensure that the main defects can be captured in a short time. This preliminary detection not only improves production efficiency, but also lays a foundation for subsequent detailed analysis.

[0063] During the defect detection process, the laser scanning system real-time obtains the defect data information on the surface of the deep holes, providing a rich data basis for subsequent preprocessing and feature extraction;

[0064] These data include key parameters such as the intensity, time, and phase difference of the reflected signal, which reflect the morphology and state of the hole wall surface.

[0065] Real-time data acquisition is the core of the entire detection process, ensuring that the system can promptly obtain changes in the surface state. Through the high-speed data acquisition module, the laser scanning system can continuously record the reflected signals of each scanning point. These data not only include the location and characteristics of significant defects but also cover potential information on subtle changes, providing a rich data basis for subsequent preprocessing and feature extraction.

[0066] Preprocess the obtained defect data information to improve data quality and the accuracy of analysis;

[0067] The preprocessing steps include noise filtering, signal smoothing, data normalization, etc., to improve data quality and the accuracy of analysis. Preprocessing is an important link to ensure that subsequent analysis steps can accurately identify key features. First, remove random noise and environmental interference in the signal through filtering algorithms (such as low-pass filtering, high-pass filtering, or band-pass filtering). Then, use signal smoothing techniques (such as moving average, weighted average) to reduce signal fluctuations, making defect features more obvious. Finally, perform data normalization to unify the scales of different signal parameters, facilitating subsequent feature extraction and model analysis.

[0068] From the preprocessed defect information, extract the key features that reflect the subtle defects on the surface of the current printed circuit board deep hole, and conduct a detailed analysis of the extracted key features under the detection window to quantify the defect degree of the surface quality of the current printed circuit board deep hole;

[0069] From the preprocessed defect information, extract the key features that reflect the subtle defects on the surface of the current printed circuit board deep hole. The extracted features include changes in the reflected signal in the defect edge region and changes in the phase of the reflected signal. Under the detection window, analyze the changes in the reflected signal in the defect edge region and the changes in the phase of the reflected signal, respectively generate a reference value for the defect edge transition and a reference value for the phase change, and quantify the defect degree of the surface quality of the current printed circuit board deep hole through the reference value for the defect edge transition and the reference value for the phase change.

[0070] A gradually changing reflection signal intensity in the defect edge region usually indicates that the surface of the deep hole in the circuit board is in a state of subtle defects. Small defects such as microcracks, bubbles, or surface non-uniformities usually do not cause a drastic change in the reflection signal intensity, but they can trigger a gradual change in the reflection signal intensity. This change usually occurs in the edge region of the defect rather than the center of the defect, so it appears as a gradually changing reflection signal. Tiny surface defects, such as cracks or bubbles, change the propagation path and scattering pattern of the light beam on the surface, resulting in a slight fluctuation in the amplitude of the reflection signal with position or angle. This subtle change in the reflection intensity is usually difficult to capture by traditional high-sensitivity detection methods because these changes are small in magnitude and may be close to the system noise level. Through fine signal analysis, especially in the edge region of the defect, these gradually changing signals can be accurately identified, thus reflecting the existence of subtle defects on the surface of the deep hole. Different from general surface irregularities, subtle defects usually show a progressive change in the reflection signal, indicating the possible existence of microscopic defects such as cracks and bubbles. Although these defects have a limited impact on the overall structure, long-term accumulation may lead to a decline in performance or failure. Therefore, the gradual change of the reflection signal in the defect edge region is an important indicator that there are subtle but potential defects on the surface of the deep hole, which deserves high attention.

[0071] The specific steps to analyze the change of the reflection signal in the defect edge region under the detection window and generate the defect edge transition reference value are as follows:

[0072] Scan the surface of the deep hole in the circuit board through a laser scanning system to obtain the reflection signal, determine the region containing the defect, and the calculation formula is as follows:

[0073]

[0074] , where R edge is the reflection signal intensity in the defect edge region, I(z, θ) is the reflection intensity signal, which represents the signal intensity reflected by the laser beam at the position z on the surface of the deep hole and the scanning angle θ, Z 1 is the starting position of the defect region, Z 2 is the ending position of the defect region, θ 1 is the starting angle of the laser scan, θ 2 is the ending angle of the laser scan;

[0075] The purpose of this step is to determine the reflection signal in the defect edge region and provide data support for subsequent change analysis.

[0076] z 1 、z 2 Determine the spatial range of the defect on the surface of the deep hole, which is usually determined by analyzing the amplitude of the signal intensity change, and identify the starting position and ending position of the defect.

[0077] θ 1 、 θ 2 Determines the angular range of the laser beam scanning, helps to determine the variation of the reflected signal at different angles, and thus identifies the details of the defect area.

[0078] After determining the defect edge area, analyze the variation trend of the reflected signal in this area, and quantify the variation amplitude by calculating the derivative of the signal change. The calculation expression is as follows:

[0079]

[0080] , where Slope edge is the slope of the change in the reflected signal within the defect edge area, measuring the degree of intensity change of the reflected signal from the starting position to the ending position of the defect edge, I(z 2 , θ 2 ) is the intensity of the reflected signal at the end position (z 2 , θ 2 ) of the defect area, I(z 1 , θ 1 ) is the intensity of the reflected signal at the starting position (z 1 , θ 1 ) of the defect area;

[0081] A smaller slope indicates that the change in the reflected signal is relatively gentle, usually corresponding to minor defects; a larger slope indicates a drastic change in the signal, usually associated with larger defects or more significant surface irregularities.

[0082] In the edge area of the surface defect of the deep hole, the change in the reflected signal is not linear. To accurately reflect the subtle degree of the defect, a non - linear mapping function is introduced to quantify the signal transition. The calculation expression is as follows:

[0083]

[0084] , where Transition exp is the defect edge transition intensity, k i is the weighting coefficient at the i - th position z i , I(z i , θ j ) refers to the intensity of the reflected signal at the i - th position z and the j - th angle θ, I 0 is the reference signal intensity, used to compare with the actually measured signal intensity to calculate the change in the signal, and n is the total number of positions;

[0085] This step weights the degree of signal change with the position and uses exponential decay to emphasize the small changes in the edge transition, adapting to the identification of subtle defects.

[0086] The reflected signal intensity R of the defect edge area edge , the slope of the reflection signal change in the defect edge area edge And the defect edge transition strength Transition exp Generate defect edge transition reference value, the generation formula is as follows:

[0087]

[0088] , where Defect Edge Transition It is the reference value of defect edge transition.

[0089] This step accurately quantifies subtle changes in defect edges by combining the signal slope with a weighted average of the nonlinear transition index.

[0090] It can be seen from the defect edge transition reference value that the smaller the defect edge transition reference value generated after analyzing the change of the reflection signal in the defect edge area under the detection window, it usually indicates that there is a subtle defect on the surface of the deep hole of the circuit board. The defect edge transition reference value quantifies the degree of gradual change of the reflection signal between the defect area and the normal area. When there is a subtle defect on the surface of the deep hole, the edge area of ​​the defect will cause a slight change transition in the reflection signal, but the amplitude of this change is small, so the transition area formed is relatively gentle. This gradual change feature causes the defect edge transition reference value to be small, indicating that the defect may be small and not obvious, and is in a subtle defect state at the microscopic level. On the contrary, if the defect is more obvious or the area is larger, the reflection signal changes more drastically, resulting in a steeper transition area at the edge of the defect, thereby generating a larger defect edge transition reference value.

[0091] When there is a slight change in the phase of the reflected signal, it usually indicates the existence of subtle defects on the surface of the deep holes in the current circuit board, especially when defects such as microcracks, bubbles, or slight surface non-uniformities occur. The change in the phase of the reflected signal is caused by the optical path difference or refractive index change when light waves pass through surface defects. Although the microcracks or bubbles on the deep hole surface may not significantly affect the reflection intensity, they will change the propagation path of light at the microscopic scale, resulting in a slight shift in the phase of the reflected light wave. This tiny phase change is difficult to detect by traditional detection methods because it is not as easy to capture as the intensity change. Therefore, the change in the phase of the reflected signal is usually a sensitive indication of the existence of subtle defects and can be detected before the surface defects have significantly expanded or had a serious impact on the overall electrical performance. In addition, the phase change is particularly sensitive to small defects because minor surface changes may only cause weak perturbations in the phase, and these changes can be accurately identified in phase-sensitive detection methods. Therefore, by detecting the change in the phase of the reflected signal, it is possible to effectively identify the tiny defects on the surface of the deep holes in the circuit board that are difficult to detect by the naked eye or cannot be captured by conventional intensity detection, thereby improving the overall defect detection ability and preventing potential quality problems.

[0092] The specific steps for analyzing the phase change of the reflected signal under the detection window to generate a phase change reference value are as follows:

[0093] For each scanning point of the laser scanning system, the phase data of the reflected signal is measured. The data usually comes from the phase difference when the laser beam irradiates the surface of the deep holes in the circuit board and is reflected back to the sensor. The formula for the phase data of the reflected signal is as follows:

[0094]

[0095] where is the phase of the reflected signal at the coordinate point (x, y), A(x, y) is the amplitude distribution of the reflected signal at the coordinate point (x, y) on the surface of the deep holes in the circuit board, ω is the angular frequency of the laser beam, t is the time variable, j is the imaginary unit, and arg is the phase extraction function;

[0096] The purpose of this process is to extract the phase information of the reflected signal at each position and provide data support for subsequent analysis.

[0097] By performing a differential analysis on the phase data of each adjacent scanning point, the local phase change on the surface of the deep hole is calculated, which is achieved by calculating the phase difference between adjacent points. The calculation expression is as follows:

[0098]

[0099] where is the phase of the reflected signal at the coordinate point (x + 1, y + 1), referring to the phase of the point adjacent to the current point (x, y). is the phase difference between adjacent points, that is, the phase difference between the coordinate points (x, y) and the adjacent coordinate point (x + 1, y + 1), which is the difference in phase between the two points and quantifies the phase fluctuation of the reflected signal between adjacent points;

[0100] This step calculates the phase difference between each pair of adjacent points in the scanning window. The local phase difference reflects the influence of minute surface changes on the propagation of light waves, especially the influence of fine defects such as microcracks and air bubbles. These minute phase changes may indicate the existence of subtle structural anomalies or defects on the surface. Especially when the change in the intensity of the reflected signal is not significant, the weak fluctuation of the phase change can more sensitively reveal the defects.

[0101] After calculating the phase difference between adjacent points the phase difference is quantified to obtain the local phase change degree. To quantify the phase change, a phase perturbation amplitude coefficient is introduced, and the calculation formula is as follows:

[0102]

[0103] , where C PD (x, y) is the local phase change degree, used to quantify the degree of minute phase change at the coordinate point (x, y), p is the adjustment exponent, controlling the sensitivity of the local phase change, is the maximum observed phase change amplitude;

[0104] By performing a weighted sum of the local phase change degree C PD (x, y), a general measure of the phase change is generated for the entire deep hole surface area.

[0105] Performing a weighted aggregation of all local phase change degrees C PD (x, y) to generate a phase change reference value, and the generation formula is as follows:

[0106] Phase Shift =∫∫ Ω exp(-αC PD (x,y))dx dy

[0107] , where Ω is the detection area, that is, the spatial area of the deep hole on the circuit board involved in the laser scanning process, α is the attenuation factor, used to control the influence degree of the local phase change degree C PD (x, y) on the phase change reference value, determining the "sensitivity" of the phase change and the response speed to minute changes, Phase Shift is the phase change reference value.

[0108] From the phase change reference value, it can be seen that the smaller the performance value of the phase change reference value generated after analyzing the phase change of the reflected signal under the detection window, generally indicates that the deep hole of the current circuit board has minor defects. The phase change reference value quantifies the amplitude of the phase change of the reflected signal. When there are minor defects on the surface of the deep hole, such as microcracks, bubbles, or slight surface unevenness, although the impact of these defects is relatively small, they will cause a slight change in the phase of the reflected light. In this case, the smaller the phase change reference value, the more subtle the phase change of the reflected signal, usually indicating the presence of minor surface defects. On the contrary, if the phase change reference value is large, it indicates that the phase fluctuation of the reflected signal is relatively obvious, usually associated with larger defects or more significant surface changes.

[0109] Input the analyzed key features into a pre-trained deep learning model, and use the deep learning model to intelligently evaluate the quality defects of the deep holes in the current circuit board;

[0110] After analyzing the extracted key features, input the generated defect edge transition reference value and phase change reference value into a pre-trained deep learning model, generate a defect type evaluation index through the deep learning model, and use the defect type evaluation index to intelligently evaluate the quality defects of the deep holes in the current circuit board.

[0111] A pre-trained deep learning model refers to, before actual application, learning and optimizing the deep learning algorithm through a large number of training data sets so that it can automatically perform predictions and classifications for specific tasks. In this example, the deep learning model is used to intelligently evaluate the quality defects on the surface of the deep holes in the circuit board, especially during the defect type evaluation process. The pre-trained model, based on the input key features, such as the defect edge transition reference value and phase change reference value, learns the relationships and patterns between these features and different defect types through the training process of a large amount of historical data. In this way, the model can not only identify defects but also accurately judge the type of defects and their possible impacts based on the different characteristics of the defects. This process usually involves multiple stages, such as data collection, feature extraction, training, and optimization, and finally obtains a model that can accurately evaluate new data.

[0112] The training process of a deep learning model generally includes two core parts: feature learning and model optimization. First, by using a labeled dataset, the model can "learn" which features are crucial for identifying different types of defects. In this solution, the model gradually optimizes its recognition ability by analyzing complex features such as defect edge transitions and phase changes. As the training progresses, the weights and parameters of the model are continuously adjusted to better adapt to new data inputs and improve the accuracy and robustness of recognition. Second, model optimization includes selecting an appropriate network architecture (such as Convolutional Neural Network CNN, Long Short-Term Memory Network LSTM, etc.), as well as setting reasonable loss functions and optimization algorithms to improve the generalization ability of the model and avoid overfitting. After training, the model will be able to efficiently evaluate different deep hole surface defects, generate a defect type evaluation index based on defect features, and thus help quickly identify and classify different defect types. This intelligent evaluation method based on deep learning greatly improves the efficiency and accuracy of defect detection, reduces manual intervention, and makes quality control more refined and automated.

[0113] The deep learning model is not limited here, and it can realize the comprehensive analysis of the defect edge transition reference value DefectEdge Transition and the phase change reference value Phase Shift to generate the defect type evaluation index DefectType Assessment Any deep learning model can be used. To implement the technical solution of the present invention, the present invention provides a specific implementation method;

[0114] The formula for generating the defect type evaluation index is as follows:

[0115]

[0116] , where z 1 , z 2 are respectively the preset proportionality coefficients of the defect edge transition reference value Defect Edge Transition and the phase change reference value Phase Shift , and z 1 , z 2 are both greater than 0.

[0117] It can be seen from the defect type evaluation index that the smaller the value of the defect edge transition reference value generated by analyzing the change of the reflection signal in the defect edge area under the detection window, and the smaller the value of the phase change reference value generated by analyzing the phase change of the reflection signal under the detection window, the smaller the value of the defect type evaluation index generated by the pre-trained deep learning model for intelligent evaluation of the current deep hole quality defect of the circuit board, which indicates that there are minor defects on the surface of the current deep hole of the circuit board, and vice versa, it indicates that the surface of the current circuit board has significant defects.

[0118] According to the evaluation results of the deep learning model, the quality defects of the current circuit board deep holes are divided into two categories: "minor defects" and "significant defects";

[0119] The defect type evaluation index generated by the pre-trained deep learning model when the deep hole quality defects of the current circuit board are intelligently evaluated is compared with the pre-set defect type evaluation index reference threshold, and the deep hole quality defects of the current circuit board are divided. The division steps are as follows:

[0120] If the defect type evaluation index is greater than or equal to a preset defect type evaluation index reference threshold, the current circuit board deep hole quality defect is classified as a minor defect;

[0121] If the defect type evaluation index is less than a preset defect type evaluation index reference threshold, the current circuit board deep hole quality defect is classified as a significant defect.

[0122] Minor defects usually refer to those defects that appear as tiny and imperceptible defects on the surface of deep holes. The physical size of these defects is small and usually does not cause significant changes in reflection signals or phase fluctuations. Therefore, they are difficult to identify in traditional detection methods. Significant defects refer to those defects that can significantly affect the performance and reliability of circuit boards. They usually show obvious physical changes on the surface and can be captured in time during detection through significant changes in reflection signals, phase fluctuations or changes in geometric shape.

[0123] For significant defects, continue to perform quality inspections at preset sensitivity to ensure rapid detection and correction to avoid production delays;

[0124] For significant defects, continue to perform quality inspections at the preset sensitivity to ensure rapid detection and correction, and avoid production delays in order to ensure the efficiency and stability of the inspection process. Significant defects are usually manifested as large or obvious changes in reflection signals, and the fixed preset sensitivity is sufficient to effectively identify these defects. Therefore, when detecting these defects, continuing to use the preset sensitivity can quickly locate and mark the defective area, avoiding unnecessary over-analysis or redundant operations, thereby speeding up the inspection. In addition, rapid detection and correction of significant defects can prevent defects from being missed or delayed during the production process, thereby affecting subsequent production links, ensuring smooth production processes and stable product quality. In this way, the system can maximize production efficiency while ensuring detection accuracy, and reduce production stagnation or repair delays caused by detection lags.

[0125] For subtle defects, the detection sensitivity is dynamically improved based on the evaluation results of the deep learning model, and the current deep hole defects are re-identified with higher sensitivity;

[0126] For minor defects, based on the evaluation results of the deep learning model, the detection sensitivity is dynamically increased. The specific steps to re-identify the current deep hole defects with higher sensitivity are as follows:

[0127] For the areas determined to be minor defects, the detection sensitivity is dynamically adjusted according to the current defect information. Based on the deep learning evaluation results, the sensitivity adjustment factor is calculated. This factor will dynamically adjust the detection sensitivity based on the defect type evaluation index to better identify minor defects. The calculation expression is as follows:

[0128]

[0129] , where SAF is the sensitivity adjustment factor, Defect Type ref is the reference threshold of the defect type evaluation index, Defect Type Assessment is the defect type evaluation index, μ is the factor controlling the exponential growth, used to adjust the response intensity of the sensitivity factor, γ is the attenuation factor, indicating the influence of signal changes on the sensitivity, especially playing a role in minor defects with small changes in the reflection signal amplitude, ΔSignal is the change amplitude of the reflection signal on the deep hole surface, reflecting the influence of the defect on the signal;

[0130] Through the above steps, the sensitivity adjustment factor SAF can adjust the sensitivity according to the severity of the defect and the signal changes, enabling the system to provide sufficient sensitivity when facing minor defects while avoiding over-detection of normal areas.

[0131] Based on the sensitivity adjustment factor SAF, the sensitivity is reset and the quality inspection of the current deep hole surface of the circuit board is carried out. The calculation expression is as follows:

[0132]

[0133] , where S preset is the preset sensitivity, η is the weighting factor for sensitivity adjustment, determining the influence intensity of defect severity on sensitivity adjustment, δ is the flexible factor controlling sensitivity adjustment, used to balance the adjustment amplitude and prevent over-adjustment, ζ is the non-linear index of sensitivity adjustment, which controls the non-linear degree of sensitivity adjustment to ensure that the sensitivity adjustment does not exceed a reasonable range, S adjusted is the adjusted sensitivity.

[0134] Through the above steps, the sensitivity can be increased when targeting minor defects, thus capturing those small changes that may be missed by traditional detection methods. This dynamic adjustment process enables the detection system to ensure the overall production efficiency while not missing any minor defects, improving the stability of product quality.

[0135] Through dynamic sensitivity adjustment and intelligent evaluation based on deep learning, this solution significantly improves the detection accuracy of micro-defects. In traditional laser scanning systems, the fixed sensitivity cannot effectively identify subtle defects, leading to missed detections. However, this method can capture even micro-cracks, bubbles, or slight non-uniformities with extremely small variations in reflected signals by evaluating in real time and dynamically adjusting the sensitivity. By extracting key features reflecting subtle defects (such as changes in reflected signals and phase changes at the defect edges) and combining them with a deep learning model for precise evaluation, the system can greatly reduce the missed detection rate and ensure the reliability and stability of high-frequency and high-performance circuit boards.

[0136] While ensuring high-precision detection, this invention also effectively balances production efficiency and quality control. The initial sensitivity setting is used for quickly screening obvious defects, avoiding production delays caused by over-detection. For subtle defects, the system dynamically increases the sensitivity according to the deep learning evaluation results, thus accurately identifying those defects missed by traditional methods while avoiding bottlenecks in the production process caused by over-detection. In this way, the production line can not only operate efficiently but also ensure the quality and long-term reliability of the final product. Especially in high-demand application environments, it can effectively prevent potential electrical failures or signal interferences.

[0137] The above formulas are all dimensionless and take their numerical values for calculation. The formula is obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters in the formula are set by those skilled in the art according to the actual situation.

[0138] Only some exemplary embodiments of the present invention have been described by way of illustration. Undoubtedly, for those of ordinary skill in the art, the described embodiments can be modified in various different ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0139] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising one..." does not exclude the existence of additional identical elements in the process, method, article or device comprising the element.

[0140] It should be understood that in various embodiments of the present application, the magnitudes of the serial numbers of the above processes do not imply the sequence of execution, and the execution sequence of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0141] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0142] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0143] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0144] In addition, the functional units in various embodiments of the present application can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.

[0145] As described above, only the specific implementation manners of the present application are provided, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present application, and all of them should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0146] Only some exemplary embodiments of the present invention have been described by way of illustration above. Undoubtedly, for those of ordinary skill in the art, the described embodiments can be modified in various different ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the protection scope of the claims of the present invention.

Claims

1. A method for controlling the surface uniformity of deep holes in a circuit board, characterized in that: The following steps are involved: The laser scanning system performs defect detection on the deep hole surface of the circuit board to be inspected with a preset sensitivity, quickly screening out defective areas while ensuring overall inspection efficiency; During defect detection, the laser scanning system acquires defect data information on the deep hole surface in real time, providing a rich data basis for subsequent preprocessing and feature extraction; Pre-process the acquired defect data information to improve data quality and analysis accuracy; From the pre-processed defect information, the key features reflecting the subtle defects on the deep hole surface of the current circuit board are extracted, and the extracted key features are analyzed in detail under the detection window to quantify the defect degree of the deep hole surface quality of the current circuit board; The analyzed key features are input into the pre-trained deep learning model, and the deep learning model is used to intelligently evaluate the quality defects of the deep holes in the current circuit board; According to the evaluation results of the deep learning model, the quality defects of the current circuit board deep holes are divided into two categories: "minor defects" and "significant defects"; For significant defects, continue to perform quality inspections at preset sensitivity to ensure rapid detection and correction to avoid production delays; For subtle defects, the detection sensitivity is dynamically improved based on the evaluation results of the deep learning model, and the current deep hole defects are re-identified with higher sensitivity.

2. A circuit board deep hole surface uniformity control method according to claim 1, characterized in that: From the preprocessed defect information, key features reflecting subtle defects on the surface of the deep holes of the current circuit board are extracted. The extracted features include changes in reflection signals and phases of reflection signals in the defect edge area. Under the detection window, changes in reflection signals and phases of reflection signals in the defect edge area are analyzed to generate defect edge transition reference values ​​and phase change reference values, respectively. The defect degree of the surface quality of the deep holes of the current circuit board is quantified by the defect edge transition reference values ​​and phase change reference values.

3. A circuit board deep hole surface uniformity control method according to claim 2, characterized in that: The specific steps for analyzing the change of the reflection signal in the defect edge area under the detection window to generate the defect edge transition reference value are as follows: The laser scanning system is used to scan the surface of the deep hole of the circuit board, obtain the reflection signal, and determine the area containing the defect. The calculation expression is as follows: ， In the formula, R edge is the reflection signal intensity of the defect edge area, I(z, θ) is the reflection intensity signal, which indicates the signal intensity reflected by the laser beam at the deep hole surface position z and scanning angle θ, z1 is the starting position of the defect area, z2 is the ending position of the defect area, θ1 is the starting angle of the laser scanning, and θ2 is the ending angle of the laser scanning; After determining the defect edge area, analyze the change trend of the reflected signal in this area, and quantify the change amplitude by calculating the derivative of the signal change. The calculation expression is as follows: ， Where Slope edge is the slope of the reflection signal change in the defect edge area, I(z2, θ2) is the reflection signal intensity at the end position (z2, θ2) of the defect area, and I(z1, θ1) is the reflection signal intensity at the start position (z1, θ1) of the defect area; In the edge area of ​​deep hole surface defects, the change of reflection signal is not linear. In order to accurately reflect the subtlety of defects, a nonlinear mapping function is introduced to quantify the transition of the signal. The calculation expression is as follows: ， In the formula, Transition exp is the defect edge transition strength, k i is the i-th position z i The weighting coefficient, I(z i ,θ j ) refers to the reflected signal strength at the i-th position z and the j-th angle θ, I0 is the reference signal strength, and n is the total number of positions; The reflected signal intensity R of the defect edge area edge , the slope of the reflection signal change in the defect edge area edge And the defect edge transition strength Transition exp Generate defect edge transition reference value, the generation formula is as follows: ， Where, Defect Edge Transition It is the reference value of defect edge transition.

4. A circuit board deep hole surface uniformity control method according to claim 2, characterized in that: The specific steps for analyzing the phase change of the reflected signal in the detection window to generate a phase change reference value are as follows: The laser scanning system measures the reflected signal phase data for each scanning point. The reflected signal phase data formula is as follows: ， In the formula, is the phase of the reflected signal at the coordinate point (x, y), A(x, y) is the amplitude distribution of the reflected signal at the coordinate point (x, y) on the surface of the deep hole of the circuit board, ω is the angular frequency of the laser beam, t is the time variable, j is the imaginary unit, and arg is the phase extraction function; By performing a difference analysis on the phase data of each adjacent scanning point, the local phase change on the deep hole surface is calculated. This is achieved by calculating the phase difference between adjacent points. The calculation expression is as follows: ， In the formula, is the phase of the reflected signal at the coordinate point (x+1, y+1), which refers to the phase of the point adjacent to the current point (x, y). is the phase difference between adjacent points, that is, the phase difference between the coordinate point (x, y) and the adjacent coordinate point (x+1, y+1), that is, the phase difference between the two points; Calculating the phase difference between adjacent points After that, the phase difference is quantified to obtain the local phase change. In order to quantify the phase change, the phase disturbance amplitude coefficient is introduced. The calculation formula is as follows: ， In the formula, C PD (x, y) is the local phase change, p is the adjustment index, is the maximum observed phase change amplitude; For all local phase changes C PD (x, y) is weighted and summarized to generate a phase change reference value. The generation formula is as follows: Phase Shift =∫∫ Ω exp(−αC PD (x,y))dx dy, Where Ω is the detection area, α is the attenuation factor, Phase Shift is the phase change reference value.

5. A circuit board deep hole surface uniformity control method according to claim 2, characterized in that: After analyzing the extracted key features, the generated defect edge transition reference value and phase change reference value are input into the pre-trained deep learning model, and the defect type evaluation index is generated by the deep learning model. The defect type evaluation index is used to perform an intelligent evaluation of the deep hole quality defects of the current circuit board.

6. A circuit board deep hole surface uniformity control method according to claim 5, characterized in that: The defect type evaluation index generated by the pre-trained deep learning model when the deep hole quality defects of the current circuit board are intelligently evaluated is compared with the pre-set defect type evaluation index reference threshold, and the deep hole quality defects of the current circuit board are divided. The division steps are as follows: If the defect type evaluation index is greater than or equal to a preset defect type evaluation index reference threshold, the current circuit board deep hole quality defect is classified as a minor defect; If the defect type evaluation index is less than a preset defect type evaluation index reference threshold, the current circuit board deep hole quality defect is classified as a significant defect.

7. A circuit board deep hole surface uniformity control method according to claim 6, characterized in that: For subtle defects, based on the evaluation results of the deep learning model, the detection sensitivity is dynamically improved, and the specific steps for re-identifying the current deep hole defects with higher sensitivity are as follows: For areas judged as minor defects, the detection sensitivity is dynamically adjusted according to the current defect information. Based on the deep learning evaluation results, the sensitivity adjustment factor is calculated. The calculation expression is as follows: ， Where SAF is the sensitivity adjustment factor, Defect Type ref is the reference threshold of the defect type evaluation index, DefectType Assessment is the defect type evaluation index, μ is the factor controlling the exponential growth, γ is the attenuation factor, and ΔSignal is the variation amplitude of the deep hole surface reflection signal; Based on the sensitivity adjustment factor SAF, the sensitivity is reset and the quality inspection of the deep hole surface of the current circuit board is performed. The calculation expression is as follows: ， In the formula, S preset is the preset sensitivity, η is the weighting factor of sensitivity adjustment, δ is the flexibility factor of sensitivity adjustment, ζ is the nonlinear index of sensitivity adjustment, S adjusted is the adjusted sensitivity.

Images