Vacuum-etching-based PCB fine line side-etching inhibition method

Abstract

The application relates to a PCB precision line side etching inhibition method based on vacuum etching, which comprises the following steps: acquiring line feature data from a printed circuit board design file, constructing a multi-dimensional parameter matrix containing coordinate positions and feature values; establishing a calculation model of etching liquid flow rate by using a regression analysis algorithm according to the line feature variables in the multi-dimensional parameter matrix; generating the flow rate distribution of each region through the calculation model, and adjusting the flow rate value in combination with differential control parameters; generating a flow rate control instruction sequence according to the adjusted flow rate value, driving the etching equipment to adjust the etching liquid flow rate of each nozzle; acquiring deviation data of the current flow rate and the target flow rate through real-time monitoring, adjusting the flow regulating valve to make the flow rate converge to the target range; collecting line edge shape data in the etching process, calculating the side etching degree and updating the flow rate parameters according to the result; and optimizing the regression coefficients in the calculation model by associating data, and establishing an adaptive parameter optimization mechanism.

Description

Technical Field

[0001] This application relates to the field of etching technology, specifically to a method for suppressing side etching of precision PCB circuits based on vacuum etching. Background Technology

[0002] Traditional etching processes rely mainly on experience-based adjustments and single-parameter control, making it difficult to adapt to the complex and ever-changing characteristics of circuits. Existing methods generally suffer from coarse etching parameter settings and an inability to perform fine control over different areas, leading to frequent side etching and severely impacting product yield.

[0003] The core challenge we face is that there is no precise mechanism to establish a correspondence between the etching solution flow rate and the circuit features. Different combinations of line width, copper foil thickness and line density form a complex multi-dimensional parameter space. Existing technologies cannot accurately identify and quantify the interaction between these parameters. This inaccuracy in parameter matching further leads to the blindness of the flow rate control strategy during the etching process. Since it is impossible to pre-determine differentiated control schemes based on the specific design features of the printed circuit board, the etching equipment can only use a uniform flow rate parameter to process the entire board surface, ignoring the differentiated requirements of different areas for etching conditions. This one-size-fits-all approach results in over-etching in some areas and insufficient etching in others, ultimately leading to uncontrollable side etching defects.

[0004] Establishing a precise matching relationship between etching solution flow rate and multidimensional circuit characteristic parameters, and developing an intelligent system that can automatically generate regional flow rate control strategies based on design documents, has become a key issue in solving the problem of suppressing side etching of precision circuits during vacuum etching. Summary of the Invention

[0005] In order to solve the problems existing in the prior art, the purpose of this application is to provide a method for suppressing side etching of precision PCB circuits based on vacuum etching.

[0006] The method for suppressing side etching of precision PCB circuits based on vacuum etching described in this application includes:

[0007] Step S101: Obtain the numerical distribution information of line width, copper foil thickness and line density as described in the printed circuit board design file, extract the geometric feature parameters of each region through image recognition algorithm, establish a multi-dimensional parameter matrix containing coordinate position and feature values, and obtain a complete line feature dataset.

[0008] Step S102: Based on the three variables W (line width), T (copper foil thickness), and D (line density) in the multidimensional parameter matrix, a calculation model V = aW + bT + cD + d for the etching solution flow rate V is established using a multivariate regression analysis algorithm, where a, b, c, and d are regression coefficients, and the values ​​of each coefficient are determined by training with historical etching data.

[0009] Step S103: If the line width is less than 0.1 mm and the line density is greater than the preset density threshold, a low flow rate control strategy is adopted to reduce the etching solution flow rate to 70% of the reference value, while extending the etching time to 1.3 times the reference time to obtain differentiated control parameters for the precision line area.

[0010] Step S104: Using the flow rate calculation model and differentiated control parameters, generate an etching solution flow rate distribution map for each region of the printed circuit board, and use an interpolation algorithm to supplement the flow rate transition values ​​between adjacent regions to obtain a continuous flow rate control command sequence.

[0011] Step S105: According to the flow rate control command sequence, drive the flow regulating valve in the etching equipment to adjust the flow rate of the etching liquid in each nozzle according to the time sequence and position coordinates, and obtain the deviation value between the current flow rate and the target flow rate through a real-time monitoring algorithm.

[0012] Step S106: If the deviation value exceeds the preset deviation threshold, the proportional-integral-derivative control algorithm is used to adjust the opening of the flow regulating valve so that the actual flow rate converges to the target flow rate range, and it is determined whether the temperature and concentration of the etching solution are within the standard range.

[0013] Step S107: The edge morphology data of the line during the etching process is collected in real time by an optical detection sensor. The edge detection algorithm is used to calculate the side etching degree. If the side etching degree exceeds the preset side etching threshold, the flow rate parameter of the corresponding area is adjusted immediately.

[0014] Step S108: Obtain the correlation data between the lateral etching degree value and the flow rate parameter, update the regression coefficient in the flow rate calculation model through machine learning algorithm, establish an adaptive parameter optimization mechanism, and obtain continuously improved etching control accuracy.

[0015] Preferably, in step S101, obtaining circuit feature data from the printed circuit board design file and constructing a multidimensional parameter matrix containing coordinate positions and feature values ​​includes:

[0016] The numerical distribution information of line width, copper foil thickness, and line density is extracted from the design documents using file parsing technology.

[0017] An image recognition algorithm is used to scan the graphic areas in the design file and extract the geometric feature parameters of each area.

[0018] The geometric feature parameters are matched with coordinate positions to construct a feature mapping table; the correlation between coordinate positions and feature values ​​is analyzed through the feature mapping table to generate a multidimensional parameter matrix.

[0019] The data integrity is verified based on the multidimensional parameter matrix. If missing or abnormal data is found, it is supplemented by interpolation method to obtain the corrected feature dataset.

[0020] The feature dataset is classified and organized by setting a preset threshold, and it is determined whether the feature values ​​of each region meet the design specifications, and the line feature analysis results are output.

[0021] Preferably, in step S102, the step of establishing a calculation model for the etching solution flow rate using a regression analysis algorithm based on the line characteristic variables in the multidimensional parameter matrix includes:

[0022] The initial dataset is constructed by obtaining data on line width, copper foil thickness, and line density, as well as the corresponding etching solution flow rate, from historical etching data.

[0023] The initial dataset is cleaned by removing missing and outlier values ​​to obtain a standard dataset;

[0024] The regression coefficients were calculated using a multiple regression analysis algorithm to determine the calculation model for the etching solution flow rate.

[0025] The prediction accuracy of the computational model is evaluated using cross-validation. If the performance index is lower than a preset threshold, the feature weights are adjusted and a new computational model is generated.

[0026] The updated calculation model is used to predict flow velocity values, and its applicability is verified by comparing them with actual flow velocities, thus determining the final calculation model.

[0027] Preferably, in step S104, generating the flow velocity distribution of each region through a calculation model and adjusting the flow velocity value in conjunction with differentiated control parameters includes:

[0028] Obtain geometric feature data for each region, and generate a preliminary flow velocity distribution using the computational model;

[0029] Based on real-time data of line width and line density, if the width is lower than the preset threshold and the density exceeds the preset threshold, a low flow rate control strategy is triggered to adjust the flow rate to the reference ratio and extend the etching time to obtain differentiated control parameters.

[0030] The regional velocity distribution is optimized using the differentiated control parameters; an interpolation algorithm is used to smooth the velocity values ​​between adjacent regions, generating a continuous velocity transition distribution; if abrupt changes occur, the interpolation parameters are adjusted, and the velocity values ​​are recalculated to obtain a smooth velocity distribution.

[0031] Preferably, in step S105, generating a flow rate control command sequence based on the adjusted flow rate value to drive the etching equipment to adjust the etching solution flow rate of each nozzle includes:

[0032] Based on the adjusted flow rate value, a flow rate control command sequence is generated, the adjustment requirements of each nozzle are analyzed, and the flow rate adjustment scheme is determined.

[0033] The flow regulation scheme drives the flow regulation valve, and the valve status and flow rate change data are collected in real time to obtain the adjusted flow rate distribution.

[0034] The current dynamic value of the flow velocity is obtained through the monitoring system to determine the deviation from the target flow velocity;

[0035] If a deviation exists, the deviation data is processed using a filtering algorithm to generate a correction value;

[0036] The flow rate adjustment scheme is updated based on the correction value, the nozzle flow rate output is adjusted, the degree of matching with the target flow rate is obtained, and the dynamic change data is recorded.

[0037] Preferably, in step S106, the step of acquiring the deviation data between the current flow rate and the target flow rate through real-time monitoring and adjusting the flow regulating valve to converge the flow rate to the target range includes:

[0038] The deviation data between the current flow rate and the target flow rate is obtained by real-time monitoring. If the deviation exceeds the preset threshold, the valve opening adjustment amount is calculated by proportional-integral-derivative control algorithm.

[0039] The valve is driven to change its opening based on the adjustment amount, and real-time flow rate data is obtained.

[0040] If the real-time flow rate data converges to the target range, the temperature and concentration data of the etching solution are collected to determine whether they meet the preset standards.

[0041] If the deviation from the standard is detected, the trend of change is predicted using a regression algorithm to generate adjustment parameters.

[0042] By adjusting the parameters mentioned above, the temperature and concentration are adjusted to obtain process parameters that meet the standards, and stable data is recorded.

[0043] Preferably, in step S107, the step of collecting line edge morphology data during etching, calculating the degree of lateral etching, and updating the flow rate parameters based on the results includes:

[0044] The edge morphology data of the line is collected by an optical detection sensor to generate raw edge image data;

[0045] The original edge image data is processed using an edge detection algorithm to extract edge contour features and calculate the lateral erosion degree.

[0046] If the lateral erosion degree value exceeds a preset threshold, the area that needs adjustment is determined and an adjustment signal is generated;

[0047] The flow velocity parameter adjustment device controlled by the adjustment signal updates the flow velocity parameter of the corresponding area.

[0048] New edge morphology data were collected based on the updated flow velocity parameters, processed, and the lateral erosion degree was verified to determine the adjustment effect.

[0049] Preferably, in step S108, the step of optimizing the regression coefficients in the computational model through correlated data and establishing an adaptive parameter optimization mechanism includes:

[0050] The correlation information between the lateral etching degree data and the flow rate parameters is obtained, and the etching process is monitored through the data acquisition system to generate a structured dataset;

[0051] Feature variables were extracted and key influencing factors were identified through correlation analysis;

[0052] A flow velocity calculation model was constructed based on the key influencing factors, and an initial regression coefficient was obtained by initial training using a regression algorithm.

[0053] If the prediction error exceeds the preset threshold, the coefficients are iteratively adjusted through an optimization algorithm to obtain the optimized regression coefficients.

[0054] The calculation model is updated by updating the optimized regression coefficients, and the parameters are corrected in real time using an adaptive mechanism to obtain the updated model and determine whether the control accuracy meets the requirements.

[0055] The lateral etching suppression method for precision PCB circuits based on vacuum etching described in this application has the following advantages: it extracts circuit feature parameters from the circuit board design file through image recognition, establishes a multidimensional parameter matrix, establishes an etching solution flow rate calculation model using multivariate regression analysis, adopts a low flow rate control strategy for precision circuit areas, generates an etching solution flow rate distribution map based on the flow rate calculation model and differentiated control parameters, drives the etching equipment to adjust the flow rate of each nozzle, monitors the flow rate deviation in real time and adjusts it using a PID control algorithm, monitors the circuit edge morphology, calculates the lateral etching degree and adjusts the flow rate parameters in a timely manner, and continuously optimizes the flow rate calculation model through machine learning algorithms to achieve adaptive parameter optimization.

[0056] This invention can precisely control the flow rate of the etching solution, effectively improving the etching accuracy of printed circuit boards. It is especially suitable for differentiated etching control of precision circuits, significantly improving product yield. Attached Figure Description

[0057] Figure 1 This application describes a method for suppressing side etching of precision PCB circuits based on vacuum etching. Figure 1 ;

[0058] Figure 2This application describes a method for suppressing side etching of precision PCB circuits based on vacuum etching. Figure 2 . Detailed Implementation

[0059] like Figures 1-2 As shown, the method for suppressing side etching of precision PCB circuits based on vacuum etching described in this application includes the following steps:

[0060] Step S101: Obtain the numerical distribution information of line width, copper foil thickness and line density as described in the printed circuit board design file, extract the geometric feature parameters of each region through image recognition algorithm, establish a multi-dimensional parameter matrix containing coordinate position and feature values, and obtain a complete line feature dataset.

[0061] Step S102: Based on the three variables W (line width), T (copper foil thickness), and D (line density) in the multidimensional parameter matrix, a calculation model V = aW + bT + cD + d for the etching solution flow rate V is established using a multivariate regression analysis algorithm, where a, b, c, and d are regression coefficients, and the values ​​of each coefficient are determined by training with historical etching data.

[0062] Step S103: If the line width is less than 0.1 mm and the line density is greater than the preset density threshold, a low flow rate control strategy is adopted to reduce the etching solution flow rate to 70% of the reference value, while extending the etching time to 1.3 times the reference time to obtain differentiated control parameters for the precision line area.

[0063] Step S104: Using the flow rate calculation model and differentiated control parameters, generate an etching solution flow rate distribution map for each region of the printed circuit board, and use an interpolation algorithm to supplement the flow rate transition values ​​between adjacent regions to obtain a continuous flow rate control command sequence.

[0064] Step S105: According to the flow rate control command sequence, drive the flow regulating valve in the etching equipment to adjust the flow rate of the etching liquid in each nozzle according to the time sequence and position coordinates, and obtain the deviation value between the current flow rate and the target flow rate through a real-time monitoring algorithm.

[0065] Step S106: If the deviation value exceeds the preset deviation threshold, the proportional-integral-derivative control algorithm is used to adjust the opening of the flow regulating valve so that the actual flow rate converges to the target flow rate range, and it is determined whether the temperature and concentration of the etching solution are within the standard range.

[0066] Step S107: The edge morphology data of the line during the etching process is collected in real time by an optical detection sensor. The edge detection algorithm is used to calculate the side etching degree. If the side etching degree exceeds the preset side etching threshold, the flow rate parameter of the corresponding area is adjusted immediately.

[0067] Step S108: Obtain the correlation data between the lateral etching degree value and the flow rate parameter, update the regression coefficient in the flow rate calculation model through machine learning algorithm, establish an adaptive parameter optimization mechanism, and obtain continuously improved etching control accuracy.

[0068] like Figures 1-2 As shown, in step S101, the numerical distribution information of line width, copper foil thickness and line density described in the printed circuit board design file is obtained. Geometric feature parameters of each region are extracted by image recognition algorithm, and a multi-dimensional parameter matrix containing coordinate position and feature value is established to obtain a complete line feature dataset.

[0069] Furthermore, in step S101, raw data is extracted from the printed circuit board design file using file parsing technology, and preliminary classification is performed on line width, copper foil thickness and line density to obtain structured numerical distribution information;

[0070] Based on the structured numerical distribution information, an image recognition algorithm is used to scan the graphic areas in the design file, extract the geometric feature parameters of each area, and determine the corresponding feature description data.

[0071] By using regional analysis methods, the extracted geometric feature parameters are matched with coordinate positions to construct a feature mapping table containing location information, thereby obtaining the correlation data between features and coordinates.

[0072] Based on the feature mapping table, the correlation between coordinate position and feature value is analyzed, and matrix construction technology is used to generate a multi-dimensional parameter matrix containing line width, copper foil thickness and line density to obtain a complete set of line features;

[0073] By using a multidimensional parameter matrix, the distribution characteristics of line width, copper foil thickness and line density are verified. If data is missing or abnormal, it is supplemented by interpolation method to obtain a corrected feature dataset.

[0074] Based on the corrected feature dataset, the line feature set is classified and organized in combination with the preset threshold, and it is determined whether the feature values ​​of each area meet the design specifications to obtain the line feature analysis results.

[0075] By analyzing real-time data collected on line width and line density, and combining the results of line characteristic analysis, we can obtain the basis for judging initial process conditions and determine the direction of process parameter adjustment.

[0076] If the collected line width is lower than the preset small threshold and the line density exceeds the preset density threshold, the flow rate control strategy is triggered. The support vector machine model is used to optimize the low flow rate value and etching time to obtain the adjusted process control parameters.

[0077] Based on the adjusted process control parameters, an etching process execution plan is generated for the precision circuit area, and the specific conditions for process execution are determined.

[0078] Specifically, in step S101, raw data is extracted from the Gerber or ODB++ format of the printed circuit board design file using file parsing technology. Regular expressions are used to match the numerical ranges of line width, copper foil thickness, and line density. For example, the line width is extracted from 0.05mm to 0.5mm, the copper foil thickness is extracted from 1oz to 3oz, and the line density is calculated by the total length of the lines per unit area. After preliminary classification, a structured JSON format numerical distribution information is generated.

[0079] Based on the structured data, the edge detection algorithm in the OpenCV library is used to scan the graphic area in the design file. Combined with Hough transform, the straight line segment features are extracted, and the geometric feature parameters of each region are calculated. For example, the standard deviation of the line width is controlled within ±0.01mm, and the average value of the copper foil thickness area is calculated with an accuracy of 0.1μm, generating feature description data.

[0080] By using the region analysis method, the extracted geometric feature parameters are matched with the coordinate system of the design file, and a feature mapping table is constructed using a quadtree spatial index. For example, regions with a line width of less than 0.1mm are marked as high-precision regions and associated with coordinates (x1,y1)-(x2,y2).

[0081] Based on the feature mapping table, a three-dimensional parameter matrix is ​​constructed using the NumPy library. The first dimension stores the coordinate position, the second dimension stores the line width and copper thickness data, and the third dimension stores the line density value. For example, the matrix dimension is set to 1000×1000×3 to obtain a complete set of line features.

[0082] Data verification is performed using a multidimensional parameter matrix. If missing line width data is found in a certain area, bilinear interpolation is used to supplement the missing values ​​based on the data of four adjacent coordinate points. Abnormal values ​​of copper foil thickness are corrected by median filtering to obtain the corrected feature dataset.

[0083] Based on the corrected dataset, the preset line width tolerance threshold is ±10%, copper thickness tolerance is ±5%, and density threshold is set to 80%. The K-means clustering algorithm is used to classify the line features, determine whether each region meets the design specifications, and output classification labels.

[0084] By performing sliding window analysis on the real-time acquired linewidth data stream, with the window size set to 50 sampling points, and combining the line feature analysis results, if the average linewidth of three consecutive windows is lower than 0.08mm and the density exceeds 85%, it is determined that process adjustment is required.

[0085] When the flow rate control strategy is triggered, the input parameters of the SVM model include the current flow rate of 2.5 m / s and the etching time of 120 s. After optimization by kernel function calculation, the suggested flow rate is 1.8 m / s and the etching time is 150 s. If the deviation of the parameters output by the model exceeds ±5%, the calculation is re-iterated.

[0086] Based on the optimized parameters, an etching process execution plan is generated, with specific conditions including a flow rate controlled at 1.8±0.1m / s, an etching time set at 150±5s, and a temperature maintained at 25±1℃.

[0087] like Figures 1-2 As shown, in step S102, based on the three variables W (line width), T (copper foil thickness), and D (line density) in the multidimensional parameter matrix, a multivariate regression analysis algorithm is used to establish a calculation model for the etching solution flow rate V: V = aW + bT + cD + d, where a, b, c, and d are regression coefficients, and the values ​​of each coefficient are determined by training with historical etching data.

[0088] Further, in step S102, an initial dataset containing line width W, copper foil thickness T, line density D and corresponding etching solution flow rate V is obtained through historical etching data to construct a basic dataset for model training.

[0089] Data preprocessing methods are used to clean the initial dataset, removing missing and outlier values ​​to obtain a standardized dataset.

[0090] Based on the standardized dataset, the regression coefficients a, b, c, and d are calculated using a multiple regression analysis algorithm to determine the calculation model for the etching solution flow rate V: V = aW + bT + cD + d.

[0091] The prediction accuracy of the computational model is evaluated by cross-validation, and the performance index data of the model is obtained.

[0092] If the model performance index is lower than the preset threshold, the feature weights in the dataset are adjusted, and the multivariate regression analysis is re-executed to obtain the updated computational model.

[0093] Based on the updated calculation model, input the new line width W, copper foil thickness T, and line density D, predict the corresponding etching solution flow rate V, and obtain the predicted result value.

[0094] By comparing the predicted values ​​with the actual etching solution flow rate, the applicability of the model in the production environment is verified, and the final calculation model is determined.

[0095] By using file parsing technology, raw data is extracted from printed circuit board design files and classified in combination with the line width W, copper foil thickness T, and line density D required for the final calculation model to obtain structured numerical distribution information.

[0096] Based on the structured numerical distribution information, a multidimensional parameter matrix is ​​constructed, and interpolation methods are used to supplement missing or abnormal data to obtain a corrected feature dataset.

[0097] Specifically, in step S102, 100 sets of sample data are extracted from the historical etching database, including line width W (10-100μm), copper foil thickness T (18-70μm), line density D (20%-80%) and corresponding etching solution flow rate V (1.5-5.2m / s), forming the initial dataset;

[0098] The data was cleaned using the 3σ criterion, and outliers such as flow velocity V > 4.8 m / s were removed, leaving 85 valid data sets to form a standardized dataset.

[0099] Based on the least squares method to fit the multiple regression equation, the loss function is set as the mean squared error (MSE). Through gradient descent iterative optimization, the regression coefficients a = 0.032, b = 0.015, c = 0.021, and d = 1.207 are calculated.

[0100] The model is evaluated using K-fold cross-validation (K=5), R 2 When the value of R = 0.89 is lower than the threshold of 0.92, a weight coefficient of 1.2 is applied to the line density D for retraining, and the optimized R is obtained. 2 Increased to 0.93;

[0101] With the new parameters W = 50 μm, T = 35 μm, and D = 45%, the model output predicted value V = 3.21 m / s, which has an error of 0.94% compared with the actual measured value of 3.18 m / s on the production line.

[0102] The Gerber design file was parsed, and OpenCV was used to identify the line width region. The coordinates (X,Y) were extracted to find W=48.7μm and T=32.5μm. Combined with the region pixel ratio, D=43.6% was calculated, and a 200×200 matrix was constructed.

[0103] When the T value at coordinate (120, 80) in the matrix is ​​missing, bilinear interpolation is used to complete the surrounding data at (118, 82) = 32.1 μm and (122, 78) = 33.0 μm to generate a complete feature dataset.

[0104] like Figures 1-2 As shown, in step S103, if the line width is less than 0.1 mm and the line density is greater than the preset density threshold, a low flow rate control strategy is adopted to reduce the etching solution flow rate to 70% of the reference value, while extending the etching time to 1.3 times the reference time to obtain differentiated control parameters for the precision line area.

[0105] Furthermore, in step S103, the initial process conditions are determined by analyzing the real-time data of the line width and line density of the printed circuit board.

[0106] If the line width is less than 0.1 mm and the line density is greater than the preset density threshold, a low flow rate control strategy is triggered to determine the low flow rate value and the adjusted etching time.

[0107] Based on the determined low flow rate value, the etching solution flow rate is adjusted to 70% of the reference flow rate to obtain the adjusted flow rate parameters;

[0108] By adjusting the flow rate parameters and extending the etching time by 1.3 times the reference time, differentiated control parameters for the precision circuit area are obtained.

[0109] The support vector machine model is used to optimize the differentiated control parameters and determine whether the parameter deviation exceeds the preset range.

[0110] If the optimization results show that the parameter deviation exceeds the preset range, the flow rate and etching time are fine-tuned to obtain the final process control parameters.

[0111] Based on the final process control parameters, the preliminary velocity distribution of each region is generated by combining the velocity calculation model, and the optimized regional velocity distribution is determined.

[0112] The flow velocity values ​​between adjacent regions are smoothed by an interpolation algorithm. If there are abrupt changes, the interpolation algorithm parameters are adjusted and the flow velocity values ​​in the abrupt change regions are recalculated to obtain a smooth flow velocity distribution.

[0113] Based on the smooth flow velocity distribution, a corresponding flow velocity control command sequence is generated to adjust the operating status of the etching solution spraying equipment and obtain a precise etching solution flow velocity output.

[0114] Specifically, in step S103, the line width data of the printed circuit board is collected in real time by a sensor, such as a line width of 0.08 mm measured by a laser rangefinder. At the same time, an image processing algorithm is used to calculate the number of lines per unit area. If the density reaches 5 lines / square millimeter and exceeds the preset threshold of 4 lines / square millimeter, the control strategy is triggered.

[0115] According to the preset rules, the reference flow rate of the etching solution was reduced from 10 liters / minute to 7 liters / minute, and the reference etching time was extended from 20 minutes to 26 minutes according to the time compensation formula;

[0116] The adjusted parameters are input into the support vector machine model, and the radial basis kernel function is used for optimization calculation. When the output results show that the flow rate deviation exceeds ±0.5 liters / minute or the time deviation exceeds ±1 minute, the parameters are automatically corrected to 6.8 liters / minute and 26.5 minutes, respectively.

[0117] A computational fluid dynamics model is used to generate a regional velocity distribution map. A cubic spline interpolation algorithm is used at the junction of adjacent regions. When the velocity difference between adjacent units exceeds 0.3 L / min, the interpolation node parameters are recalculated to make the transition gradient less than 0.1 L / min·mm.

[0118] Finally, a sequence of instructions containing timestamps is generated to control the opening of the solenoid valve and keep the output fluctuation of the etching solution within ±0.2 liters / minute.

[0119] like Figures 1-2 As shown, in step S104, the etching solution flow rate distribution map corresponding to each area of ​​the printed circuit board is generated through the flow rate calculation model and the differentiated control parameters. The flow rate transition values ​​between adjacent areas are supplemented by the interpolation algorithm to obtain a continuous flow rate control command sequence.

[0120] Further, in step S104, by acquiring the geometric feature data of each area of ​​the printed circuit board and the initial flow rate parameters of the etching solution, the flow rate calculation model is input to generate preliminary flow rate distribution data;

[0121] Based on the preliminary velocity distribution data and combined with differentiated control parameters, the velocity values ​​of each region are adjusted to obtain an optimized regional velocity distribution.

[0122] An interpolation algorithm is used to smooth the velocity values ​​between adjacent regions in the optimized regional velocity distribution, generating a continuous velocity transition distribution.

[0123] If there are abrupt velocity changes in the continuous velocity transition distribution, the velocity values ​​in the abrupt region can be recalculated by adjusting the interpolation algorithm parameters to obtain a smooth velocity distribution.

[0124] Based on the smooth flow velocity distribution, a corresponding flow velocity control command sequence is generated to determine the real-time flow velocity control parameters for each region.

[0125] By adjusting the operating status of the etching solution jetting equipment through real-time flow rate control parameters, a precise etching solution flow rate output can be obtained.

[0126] The real-time monitoring data of the etching solution flow rate output is obtained and compared with the flow rate control command sequence to determine whether there is a deviation.

[0127] If the deviation between the monitoring data and the control command sequence exceeds a preset threshold, the flow velocity calculation model parameters are adjusted, and the flow velocity distribution data is regenerated.

[0128] Based on the regenerated flow rate distribution data, the flow rate regulating valves in the driving etching equipment are adjusted according to the time sequence and position coordinates to obtain the flow rate output that meets the target.

[0129] Specifically, in step S104, by collecting data on line width, spacing and copper thickness of each area of ​​the printed circuit board, and combining initial flow velocity parameters such as inlet pressure of 2.5 Bar and temperature of 25°C, a flow velocity calculation model is established using the finite element analysis method, and preliminary distribution data containing 5 zones with a flow velocity range of 1.2-3.8 m / s is output.

[0130] Based on the priority weights in the differentiated control parameters (1.5 for high-precision areas and 1.0 for ordinary areas), nonlinear optimization is performed on the initial distribution to increase the flow velocity in the key areas to 4.0 m / s.

[0131] The adjacent partition boundary data are processed using a bicubic spline interpolation algorithm to generate a transitional distribution with a grid precision of 0.1 mm. If a gradient change of more than 15% is detected at a certain boundary, a local recalculation is triggered, and the interpolation node density is adjusted to 0.05 mm before a smooth curve is regenerated.

[0132] The final distribution is converted into a time-space matrix instruction sequence, containing flow rate setpoints at 200ms intervals, and the opening of the solenoid valve is adjusted by a PID controller (0-100% corresponds to 0.5-5.0m / s).

[0133] Real-time monitoring uses ultrasonic flow meter sampling (10Hz frequency). When the actual flow velocity in area B (3.2m / s) deviates from the commanded value (3.5m / s) by more than 8%, the backpropagation algorithm updates the viscosity coefficient in the model, iteratively generates a new distribution, and drives the servo motor to adjust the lateral movement speed of the nozzle array to 12cm / s.

[0134] like Figures 1-2 As shown, in step S105, according to the flow rate control command sequence, the flow regulating valve in the etching equipment is driven to adjust the etching liquid flow rate of each nozzle according to the time sequence and position coordinates, and the deviation value between the current flow rate and the target flow rate is obtained through a real-time monitoring algorithm.

[0135] Further, in step S105, according to the flow rate control command sequence, the time series and position coordinate data are parsed, and the flow regulating valve in the etching equipment is driven to adjust the flow rate of the etching liquid in each nozzle, and the preliminary valve driving state is determined.

[0136] By using a real-time monitoring algorithm, the current flow rate data of each nozzle is collected, the deviation between the current flow rate and the target flow rate is calculated, and a flow rate deviation dataset is obtained.

[0137] If the deviation value in the flow velocity deviation dataset exceeds the preset threshold, the Kalman filter algorithm is used to smooth the deviation data to obtain the optimized deviation correction value.

[0138] Based on the optimized deviation correction value, a new flow rate control command sequence is generated, the control parameters of the time series and position coordinates are updated, and a new valve drive scheme is determined.

[0139] By adjusting the opening of the flow regulating valve through a new valve drive scheme, the output flow rate of the etching solution from each nozzle is controlled, and the adjusted flow rate distribution data is obtained.

[0140] A real-time monitoring system is used to collect adjusted flow velocity distribution data, analyze the degree of matching between the current flow velocity and the target flow velocity, and obtain the flow velocity matching analysis results.

[0141] Based on the flow rate matching analysis results, update the instruction sequence database, optimize the parameters of the flow rate regulation model, and determine the continuously optimized flow rate control model;

[0142] By using an optimized flow rate control model and combining it with the geometric feature data of each region of the printed circuit board, a preliminary regional flow rate distribution is generated, and regional flow rate control parameters are obtained.

[0143] If there are abrupt changes in the regional velocity distribution, the velocity values ​​in the abrupt change areas are recalculated by locally adjusting the interpolation algorithm parameters to obtain a smooth regional velocity distribution.

[0144] Specifically, in step S105, according to the flow rate control command sequence, time series data such as [0ms, 100ms, 200ms] and position coordinate data such as [(x1, y1), (x2, y2)] are parsed, and the flow rate regulating valve is driven to adjust the etching liquid flow rate of the nozzle to the target value of 50ml / s using a PID control algorithm, and the driving state of valve opening of 60% is determined.

[0145] The flow rate data of each nozzle is collected by a real-time monitoring algorithm with a sampling period of 10ms. The deviation of the current flow rate of 48ml / s from the target value of 50ml / s is calculated to be 2ml / s, forming a matrix dataset containing all nozzle deviations.

[0146] If there are threshold terms in the deviation matrix that exceed 1 ml / s, the Kalman filter algorithm is used to smooth the data with process noise Q = 0.1 and observation noise R = 1.0, and the corrected deviation value is 1.5 ml / s.

[0147] A new instruction sequence is generated based on the correction value, and the control parameter at coordinate (x1, y1) at time 100ms is updated to an opening degree of 65%, thus forming a new driving scheme.

[0148] After implementing the new scheme, the adjusted flow rate data collected by the electromagnetic flowmeter was 49.8 ml / s, which matched the target value of 50 ml / s by 99.6%.

[0149] The matching results are written into the SQL database, and the gradient descent method is used to optimize the flow regulation model parameter α = 0.01 to establish the updated control model.

[0150] Based on the geometric data of the PCB board (300mm x 200mm), an initial distribution of flow velocity of 45ml / s in region A and 55ml / s in region B was generated through a calculation model.

[0151] A sudden change in flow rate of 10 ml / s was detected at the boundary of the region. The smoothing coefficient λ = 0.3 was adjusted using a cubic spline interpolation algorithm, and the continuous distribution of flow rate in the transition region of 48 ml / s-52 ml / s was recalculated.

[0152] like Figures 1-2 As shown in step S106, if the deviation value exceeds the preset deviation threshold, the proportional-integral-derivative control algorithm is used to adjust the opening of the flow regulating valve so that the actual flow rate converges to the target flow rate range, and it is determined whether the temperature and concentration of the etching solution are within the standard range.

[0153] Further, in step S106, if the deviation value exceeds the preset deviation threshold, the proportional-integral-derivative control algorithm is used to calculate the adjustment amount of the flow regulating valve opening to obtain the adjusted valve opening.

[0154] Based on the adjusted valve opening, the flow regulating valve is controlled by the drive module to change the opening and obtain real-time flow rate data;

[0155] If the real-time flow rate data converges to the target flow rate range, the temperature and concentration data of the etching solution are collected by the sensor to determine whether it meets the preset standard.

[0156] If the temperature or concentration of the etching solution deviates from the preset standard, a linear regression algorithm is used to analyze the historical data of temperature and concentration to obtain the predicted adjustment parameters.

[0157] Based on the predicted adjustment parameters, the heating or cooling device is adjusted through the temperature control module to obtain the adjusted etching solution temperature;

[0158] Based on the predicted parameters, the amount of chemical substances added is controlled by the concentration adjustment module to obtain the adjusted etching solution concentration.

[0159] If the adjusted etching solution temperature and concentration both meet the preset standards, the current flow rate, temperature and concentration data are stored through the data recording module to confirm that the process parameters are stable.

[0160] The data acquisition system obtains data on the degree of lateral erosion and flow velocity parameters, resulting in a structured dataset of lateral erosion degree and flow velocity parameters.

[0161] By analyzing the correlation information, feature variables were extracted from the lateral erosion degree dataset and the flow velocity parameter dataset. The correlation between the two was calculated using the Pearson correlation coefficient method to determine the key influencing factors.

[0162] Specifically, in step S106, if the deviation between the actual flow velocity and the target flow velocity exceeds 5%, the proportional-integral-derivative control algorithm (PID parameters Kp = 0.8, Ki = 0.2, Kd = 0.1) is used to calculate the valve opening adjustment amount, and the adjusted valve opening is 65%.

[0163] Based on the adjusted valve opening, the flow regulating valve is controlled by the drive module to change the opening with a response time of 0.5 seconds, and real-time flow rate data (sampling frequency 10Hz) is obtained.

[0164] If the real-time flow rate data converges to the target flow rate range (±2% error band), the temperature (set value 25℃) and concentration (set value 10%) of the etching solution are collected by the temperature sensor (accuracy ±0.1℃) and the concentration sensor (accuracy ±0.5%) to determine whether it meets the preset standards (temperature allowable deviation ±1℃, concentration allowable deviation ±0.3%).

[0165] If the etching solution temperature (26.5℃) or concentration (9.8%) deviates from the preset standard, a linear regression algorithm will be used (historical data window of 30 minutes, goodness of fit R). 2 ≥0.95) Analyze the trends of temperature and concentration changes to obtain prediction adjustment parameters (temperature correction -1.2℃, concentration correction +0.2%);

[0166] Based on the predicted adjustment parameters, the heating or cooling device is adjusted by the temperature control module using a PID algorithm (Kp=1.0, Ki=0.3, Kd=0.05) to obtain the adjusted etching solution temperature (25.3℃).

[0167] Based on the predicted adjustment parameters, the amount of chemical substance added is controlled by the concentration adjustment module in a pulse addition mode (0.05L each time) to obtain the adjusted etching solution concentration (10.1%).

[0168] If the adjusted etching solution temperature and concentration both meet the preset standards, the current flow rate (12.5 L / min), temperature (25.3℃), and concentration (10.1%) data are stored in CSV format through the data recording module to confirm that the process parameters are stable (meeting the standards for 3 consecutive samplings);

[0169] The data acquisition system acquires lateral erosion severity data (measurement accuracy ±0.01μm) and flow velocity parameter information (12.5L / min) at 1-second intervals, resulting in a structured lateral erosion severity dataset (mean 1.2μm, variance 0.05) and flow velocity parameter dataset (mean 12.5L / min, variance 0.3).

[0170] By using correlation analysis (sliding window analysis, window size 50 sets of data), feature variables (maximum lateral erosion depth, flow velocity fluctuation rate) were extracted from the lateral erosion degree dataset and the flow velocity parameter dataset. The correlation between the two was calculated using the Pearson correlation coefficient method (r = 0.82, p < 0.01), and the key influencing factor (flow velocity fluctuation rate weight 0.75) was determined.

[0171] like Figures 1-2 As shown, in step S107, the edge morphology data of the line during the etching process is collected in real time by an optical detection sensor, and the edge detection algorithm is used to calculate the side etching degree value. If the side etching degree exceeds the preset side etching threshold, the flow rate parameter of the corresponding area is immediately adjusted.

[0172] Furthermore, in step S107, the edge morphology data of the line during the etching process is acquired in real time by an optical detection sensor to generate original edge image data;

[0173] The Canny edge detection algorithm is used to process the original edge image data, extract the line edge features, and obtain the edge contour data;

[0174] The degree of lateral erosion is calculated based on the edge contour data, and a numerical value of the degree of lateral erosion is generated.

[0175] If the lateral erosion level exceeds the preset lateral erosion threshold, the area that needs adjustment is determined and an area adjustment signal is generated.

[0176] The flow velocity parameter adjustment device is controlled by the regional adjustment signal to update the flow velocity parameters of the corresponding region and obtain the adjusted flow velocity data.

[0177] Based on the adjusted flow velocity data, new line edge morphology data are collected in real time to generate updated edge image data;

[0178] The updated edge image data is processed using the Canny edge detection algorithm to extract new edge contour features and obtain the verified lateral erosion degree value.

[0179] If the verified lateral erosion degree value still exceeds the preset lateral erosion threshold, the lateral erosion degree data and flow velocity parameter information are recorded through the data acquisition system to generate a structured lateral erosion degree dataset and flow velocity parameter dataset.

[0180] By analyzing the correlation information, feature variables were extracted from the lateral erosion degree dataset and the flow velocity parameter dataset. The correlation between the two was calculated using the Pearson correlation coefficient method to determine the key influencing factors.

[0181] Specifically, in step S107, the optical detection sensor acquires grayscale images of the etched line edges at a rate of 10 frames per second, with an image resolution of 2048×1536 pixels, generating raw edge image data.

[0182] The Canny edge detection algorithm is used to perform Gaussian filtering (σ=1.5), gradient calculation (Sobel operator), non-maximum suppression, and double thresholding (low threshold 50, high threshold 150) on the image to extract continuous edge pixels and obtain sub-pixel level precision edge contour data.

[0183] The lateral erosion width is calculated based on the edge contour data. The ideal edge line is fitted using the least squares method. The actual edge deviation distance is measured. If the lateral erosion width exceeds the preset threshold of 5μm, the coordinates of the abnormal area (X1-Y1 to X2-Y2) are marked, and an area adjustment signal is generated.

[0184] The area adjustment signal triggers the flow rate controller, which increases the etching solution flow rate in the corresponding area from 1.2 m / s to 1.5 m / s, and generates an adjusted flow rate parameter table.

[0185] After adjustment, a new edge image was acquired, and the edge contour was re-extracted using the Canny algorithm. The lateral erosion width was calculated to be 4.8 μm.

[0186] If the lateral erosion width still exceeds the threshold, record the current flow velocity of 1.5 m / s and the lateral erosion width of 4.8 μm to the SQL database to form a structured dataset containing timestamps;

[0187] A correlation analysis was performed on 100 sets of historical data in the database to extract features such as flow velocity standard deviation and mean lateral erosion. The Pearson correlation coefficient r = 0.92 was calculated, and flow velocity fluctuation was identified as the key influencing factor.

[0188] like Figures 1-2 As shown, in step S108, the correlation data between the lateral etching degree value and the flow rate parameter is obtained, the regression coefficient in the flow rate calculation model is updated through machine learning algorithm, an adaptive parameter optimization mechanism is established, and the etching control accuracy is continuously improved.

[0189] Furthermore, in step S108, the etching process is monitored in real time through a data acquisition system to obtain the raw data of the lateral etching degree and flow rate parameters, thereby obtaining a structured dataset of the lateral etching degree and the flow rate parameter.

[0190] Based on the obtained lateral erosion degree dataset and flow velocity parameter dataset, the correlation information analysis method is used to extract feature variables and determine the key influencing factors between the two.

[0191] The extracted feature variables were calculated using the Pearson correlation coefficient method, and the correlation between the lateral erosion degree and the flow velocity parameter was analyzed to obtain the correlation quantification results.

[0192] Based on the correlation quantification results, a flow rate calculation model was constructed, and the model was initialized and trained using a linear regression algorithm to obtain the initial regression coefficients;

[0193] If the prediction error of the initial regression coefficients exceeds the preset threshold, the regression coefficients are iteratively optimized using the gradient descent method to obtain the optimized regression coefficients.

[0194] The flow velocity calculation model is updated by using optimized regression coefficients, and the model parameters are corrected in real time through an adaptive parameter adjustment mechanism to obtain the updated flow velocity calculation model.

[0195] The etching process is controlled in real time by an updated flow rate calculation model, etching control accuracy data is collected, and it is determined whether the accuracy meets the preset threshold requirements.

[0196] If the etching control accuracy does not meet the preset threshold requirements, return to the correlation information analysis step, re-extract feature variables and optimize model parameters to obtain an improved flow rate calculation model;

[0197] New control commands are generated based on the improved flow rate calculation model, and etching process parameters are adjusted in conjunction with real-time monitoring data to determine continuously optimized etching control accuracy data.

[0198] Specifically, in step S108, the side etching depth and corresponding nozzle flow rate of 12 regions in the etching tank are monitored in real time by the data acquisition system at a sampling frequency of 100Hz. The acquisition cycle is 5 seconds, and a structured CSV dataset containing timestamps, location coordinates, side etching micron-level measurement values ​​and flow rate L / min data is generated.

[0199] Principal component analysis was used to extract the top three feature components with a contribution rate of over 85% from the dataset as key influencing factors, including peak etch rate, average flow velocity gradient, and transient flow pressure fluctuation coefficient.

[0200] The correlation coefficient between lateral erosion depth and peak flow velocity was calculated using Pearson correlation coefficient, which was 0.78. Four strongly correlated variables were selected when the significance level was set at p<0.01.

[0201] A multiple linear regression model y = β0 + β1x1 + ... + β4x4 was constructed based on selected variables. After initialization and training using the least squares method, initial coefficients such as β0 = 2.35 and β1 = 0.63 were obtained, and the mean square error (MSE) was 4.2 μm.

[0202] When the MSE exceeds the preset threshold of 3.0 μm, the stochastic gradient descent method with a learning rate of η = 0.01 is used for 200 iterations to make the regression coefficients converge to β0' = 2.18 and β1' = 0.71, and the MSE decreases to 2.8 μm;

[0203] The model parameters are updated every 30 seconds through an online learning mechanism. The sliding window method is used to retain the most recent 100 sets of data for refitting, and the weights of variables with a β coefficient deviation of more than 5% are dynamically adjusted.

[0204] The updated model output is converted into a PWM control signal to drive 16 proportional valves to adjust the flow rate to the target value within ±0.5L / min range, and the laser velocimeter feeds back the flow rate data to the control terminal in real time;

[0205] If the standard deviation of the side etching uniformity is still >1.2μm after three consecutive adjustments, the feature re-extraction process is triggered, and the kernel density estimation method is used to re-divide the nonlinear relationship region between flow rate and etching amount.

[0206] Finally, a three-dimensional control matrix containing valve opening degree, duration, and spatial coordinates is generated and written to the PLC actuator via the OPC-UA protocol, so that the etching depth fluctuation range is controlled within ±0.8μm.

[0207] For those skilled in the art, various other corresponding changes and modifications can be made based on the technical solutions and concepts described above, and all such changes and modifications should fall within the protection scope of the claims of this application.

Claims

1. A vacuum-etching-based fine line side-etching inhibition method for a PCB, characterized by, The application relates to a method for dynamically adjusting etching liquid flow rate in a printed circuit board (PCB) etching process. The method comprises the following steps: Obtaining line feature data from a PCB design file and constructing a multi-dimensional parameter matrix containing coordinate positions and feature values; Using a regression analysis algorithm to establish a calculation model of etching liquid flow rate based on line feature variables in the multi-dimensional parameter matrix; Generating flow rate distribution of each region through the calculation model and adjusting flow rate values in combination with differential control parameters, which specifically comprises the following steps: Triggering a low flow rate control strategy when the line width is less than 0.1 mm and the line density is greater than a preset density threshold; Reducing the etching liquid flow rate to 70% of the baseline value and extending the etching time to 1.3 times of the baseline time; Optimizing the flow rate value and etching time through a support vector machine model and reiterating the calculation when the output parameter deviation exceeds + / - 5%; Generating a flow rate control instruction sequence based on the adjusted flow rate value to drive the etching equipment to adjust the etching liquid flow rate of each nozzle; Adjusting the flow regulating valve to make the flow rate converge to the target range by monitoring the deviation data of the current flow rate and the target flow rate in real time; Collecting line edge shape data during the etching process, calculating the side etching degree and updating the flow rate parameter according to the result; 2. The method of claim 1, wherein the vacuum-etching-based fine line side-etching suppression method is characterized by, Optimizing the regression coefficients in the calculation model through associated data to establish an adaptive parameter optimization mechanism. The multi-dimensional parameter matrix comprises the following steps: Extracting the numerical distribution information of line width, copper foil thickness and line density from the Gerber format or ODB++ format of the PCB design file; Extracting geometric feature parameters of each region through an edge detection algorithm and Hough transformation; Matching the geometric feature parameters with the coordinate positions to generate a feature mapping table and constructing a parameter matrix containing three-dimensional data of line width, copper foil thickness and line density, wherein the matrix dimension is 1000x1000x3; 3. The method of claim 1, wherein the vacuum-etching-based fine line side-etching suppression method is characterized by, When there is data missing, using a bilinear interpolation method to supplement the missing values based on the data of adjacent coordinate points. The flow rate control instruction sequence comprises the following steps: Generating a preliminary flow rate distribution map of each region based on the flow rate calculation model and differential control parameters; Using a bicubic spline interpolation algorithm to supplement the flow rate transition values between adjacent regions with a grid precision of 0.1 mm; When it is detected that the flow rate gradient change of adjacent regions exceeds 15%, adjusting the interpolation node density to 0.05 mm to recalculate; 4. The method of claim 1, wherein the method is a vacuum-etching based PCB fine line side-etching inhibition method, characterized in that, Converting the final flow rate distribution into an instruction sequence matrix containing time stamps and spatial coordinates, and the time interval is 200 ms. The method for adjusting the flow regulating valve to make the flow rate converge to the target range comprises the following steps:

5. The method of claim 1, wherein the method is a vacuum-etching based PCB fine line side-etching inhibition method, characterized in that, When the flow rate deviation exceeds 5%, using a proportional-integral-derivative control algorithm to adjust the valve opening, wherein the proportional coefficient Kp=0.8, the integral coefficient Ki=0.2 and the differential coefficient Kd=0.

1. The method for calculating the side etching degree and updating the flow rate parameter according to the result comprises the following steps: Collecting line edge gray scale images at a rate of 10 frames per second through an optical detection sensor and calculating the side etching width through a Canny edge detection algorithm; if the side etching width exceeds 5 mu m, the flow rate of the region is increased by 0.3 m / s.

Images