PCB precision circuit lateral erosion inhibition method based on vacuum etching

Abstract

The invention relates to a PCB precise circuit lateral erosion inhibition method based on vacuum etching, and the method comprises the steps: obtaining circuit feature data from a printed circuit board design file, and constructing a multi-dimensional parameter matrix containing coordinate positions and feature values; according to the line characteristic variables in the multi-dimensional parameter matrix, establishing a calculation model of the flow velocity of the etching liquid by adopting a regression analysis algorithm; generating the flow velocity distribution of each area through a calculation model, and adjusting the flow velocity value in combination with the differential control parameters; generating a flow velocity control instruction sequence according to the adjusted flow velocity value, and driving the etching equipment to adjust the flow velocity of the etching liquid of each nozzle; obtaining deviation data of the current flow velocity and the target flow velocity through real-time monitoring, and adjusting a flow adjusting valve to enable the flow velocity to converge to a target range; line edge form data in the etching process are collected, the lateral erosion degree is calculated, and the flow velocity parameter is updated according to a result; and a self-adaptive parameter optimization mechanism is established through a regression coefficient in a correlation data optimization calculation model.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of etching, in particular to a PCB precision circuit side etching inhibition method based on vacuum etching. BACKGROUND

[0002] The traditional etching process mainly relies on experience adjustment and single parameter control, which is difficult to adapt to the complex and variable line feature requirements. The existing methods generally have problems such as rough etching parameter setting and inability to fine control different areas, which leads to frequent side etching phenomenon, seriously affecting product qualification rate.

[0003] The core challenge currently faced is the lack of accurate corresponding relationship establishment mechanism between etching liquid flow rate and line features. Different line widths, copper foil thicknesses and line densities form a complex multi-dimensional parameter space, and the existing technology cannot accurately identify and quantify the interaction rules between these parameters. This inaccuracy in parameter matching further leads to the blindness of flow rate control strategy in the etching process. Since it is impossible to develop differentiated control schemes in advance according to the specific design features of the printed circuit board, the etching equipment can only use uniform flow rate parameters to process the entire board, ignoring the differentiated needs of different areas for etching conditions. This one-size-fits-all approach results in over-etching in some areas, while other areas are not etched sufficiently, ultimately forming uncontrollable side etching defects.

[0004] How to establish an accurate matching relationship between etching liquid flow rate and multi-dimensional line feature parameters, and develop an intelligent system that can automatically generate regionalized flow rate control strategies based on design files, has become a key problem in solving the precision line side etching inhibition in the vacuum etching process. SUMMARY

[0005] In order to solve the problems existing in the prior art, the present application aims to provide a PCB precision circuit side etching inhibition method based on vacuum etching.

[0006] The PCB precision circuit side etching inhibition method based on vacuum etching described in the present application comprises:

[0007] Step S101, obtaining the numerical distribution information of the line width, copper foil thickness and line density in the printed circuit board design file, extracting the geometric feature parameters of each region through image recognition algorithm, establishing a multi-dimensional parameter matrix containing coordinate position and feature value, and obtaining a complete line feature data set;

[0008] Step S102, according to the line width W, copper foil thickness T and line density D in the multi-dimensional parameter matrix, a calculation model V=aW+bT+cD+d of etching liquid flow rate V is established by using multivariate regression analysis algorithm, wherein a, b, c and d are regression coefficients, and the numerical values of each coefficient are determined by training historical etching data;

[0009] 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 liquid flow rate to 70% of the reference value, and the etching time is extended to 1.3 times the reference time, to obtain the differential control parameters of the precision line region.

[0010] In step S104, the flow rate distribution map corresponding to each region of the printed circuit board is generated by the flow rate calculation model and the differential control parameters, and the interpolation algorithm is used to supplement the flow rate transition values between adjacent regions to obtain a continuous flow rate control instruction sequence.

[0011] In step S105, according to the flow rate control instruction 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 of the current flow rate from the target flow rate is obtained through the real-time monitoring algorithm.

[0012] 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 judged whether the etching liquid temperature and concentration are within the standard range.

[0013] In step S107, the line edge shape data in the etching process is collected in real time by the optical detection sensor, and the side etching degree value is calculated by the edge detection algorithm. If the side etching degree exceeds the preset side etching threshold, the flow rate parameters of the corresponding region are immediately adjusted.

[0014] In step S108, the correlation data of the side etching degree value and the flow rate parameter is obtained, the regression coefficient in the flow rate calculation model is updated by the machine learning algorithm, the adaptive parameter optimization mechanism is established, and the continuously improved etching control precision is obtained.

[0015] Preferably, in step S101, the line feature data is obtained from the printed circuit board design file, and a multi-dimensional parameter matrix containing coordinate positions and feature values is constructed, including:

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

[0017] The image recognition algorithm is used to scan the graphic region in the design file to extract the geometric feature parameters of each region;

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

[0019] According to the multi-dimensional parameter matrix, data integrity is checked, if missing or abnormal, it is supplemented by interpolation method, and a modified feature data set is obtained;

[0020] The feature data set is classified and arranged by a preset threshold value, whether the regional feature values meet the design specification is judged, and a line feature analysis result is output.

[0021] Preferably, in step S102, a regression analysis algorithm is used to establish a calculation model of etching liquid flow rate according to the line feature variables in the multi-dimensional parameter matrix, including:

[0022] The line width, copper foil thickness and line density and their corresponding etching liquid flow rate data are obtained from historical etching data to construct an initial data set;

[0023] The initial data set is cleaned to eliminate missing values and abnormal values, and a standard data set is obtained;

[0024] The regression coefficients are calculated by a multiple regression analysis algorithm to determine the etching liquid flow rate calculation model;

[0025] The prediction accuracy of the calculation model is evaluated by a cross-validation method, and if the performance index is lower than a preset threshold value, the feature weight is adjusted to regenerate an updated calculation model;

[0026] The flow rate value is predicted by the updated calculation model, and compared with the actual flow rate to verify the applicability, and the final calculation model is determined.

[0027] Preferably, in step S104, the flow rate distribution of each region is generated by the calculation model, and the flow rate value is adjusted combined with the differential control parameter, including:

[0028] The geometric feature data of each region is obtained, and the preliminary flow rate distribution is generated by the calculation model;

[0029] According to the real-time data of line width and line density, if the width is lower than a preset threshold value and the density exceeds a preset threshold value, a low flow rate control strategy is triggered, the flow rate is adjusted to a reference ratio and the etching time is prolonged, and the differential control parameter is obtained;

[0030] The regional flow rate distribution is optimized by the differential control parameter; the flow rate values between adjacent regions are smoothed by an interpolation algorithm to generate a continuous flow rate transition distribution; if there is a mutation point, the interpolation parameter is adjusted, the flow rate value is recalculated, and a smooth flow rate distribution is obtained

[0031] Preferably, in step S105, the flow rate control instruction sequence is generated according to the adjusted flow rate value to drive the etching equipment to adjust the etching liquid flow rate of each nozzle, including:

[0032] Generate flow rate control instruction sequence according to the adjusted flow rate value, analyze the adjustment requirements of each nozzle, and determine the flow regulation scheme;

[0033] Drive the flow regulation valve through the flow regulation scheme, collect valve state and flow rate change data in real time, and obtain the adjusted flow rate distribution;

[0034] Obtain the current flow rate dynamic value through the monitoring system, and judge the deviation from the target flow rate;

[0035] If there is a deviation, process the deviation data through a filtering algorithm to generate a correction value;

[0036] Update the flow regulation scheme according to the correction value, adjust the nozzle flow rate output, obtain the matching degree with the target flow rate, and record the dynamic change data.

[0037] Preferably, in step S106, the deviation data of the current flow rate and the target flow rate is obtained through real-time monitoring, and the flow regulation valve is adjusted to converge the flow rate to the target range, comprising:

[0038] Obtain the deviation data of the current flow rate and the target flow rate through real-time monitoring, and if the deviation exceeds the preset threshold, calculate the valve opening adjustment amount through a proportional-integral-derivative control algorithm;

[0039] Drive the valve to change the opening according to the adjustment amount, and obtain real-time flow rate data;

[0040] If the real-time flow rate data converges to the target range, collect etching liquid temperature and concentration data, and judge whether it meets the preset standard;

[0041] If it deviates from the standard, predict the trend of change through a regression algorithm, and generate adjustment parameters;

[0042] Adjust the temperature and concentration through the adjustment parameters, obtain process parameters that meet the standard, and record stable data.

[0043] Preferably, in step S107, the line edge shape data in the etching process is collected, the side etching degree is calculated, and the flow rate parameters are updated according to the results, comprising:

[0044] Collect line edge shape data through an optical detection sensor to generate original edge image data;

[0045] Process the original edge image data using an edge detection algorithm to extract edge contour features and calculate side etching degree values;

[0046] If the side etching degree value exceeds the preset threshold, determine the adjustment area and generate an adjustment signal;

[0047] The flow rate parameter adjusting device is controlled by the adjustment signal to update the flow rate parameter of the corresponding area;

[0048] New edge shape data is collected according to the updated flow rate parameter, and the side etching degree value is verified after processing to determine the adjustment effect.

[0049] Preferably, in step S108, the regression coefficients in the calculation model are optimized by associating data to establish an adaptive parameter optimization mechanism, comprising:

[0050] Obtain the correlation information between the side etching degree data and the flow rate parameter, monitor the etching process through the data acquisition system, and generate a structured data set;

[0051] Determine the key influencing factors by correlation analysis and extracting characteristic variables;

[0052] According to the key influencing factors, a flow rate calculation model is constructed, and a regression algorithm is used for initialization training to obtain initial regression coefficients;

[0053] If the prediction error exceeds the preset threshold, the coefficient is iteratively adjusted by an optimization algorithm to obtain optimized regression coefficients;

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

[0055] The PCB precision line side etching suppression method based on vacuum etching has the advantages that the line feature parameters in the circuit board design file are extracted by image recognition, a multi-dimensional parameter matrix is established, a multi-element regression analysis is used to establish an etching liquid flow rate calculation model, a low flow rate control strategy is used for the precision line area, an etching liquid flow rate distribution map is generated according to the flow rate calculation model and the differential control parameters, the etching equipment is driven to adjust the flow rate of each nozzle, the flow rate deviation is monitored in real time and adjusted by using a PID control algorithm, the line edge shape is monitored, the side etching degree is calculated and the flow rate parameter is adjusted in time, the flow rate calculation model is continuously optimized by using a machine learning algorithm, and adaptive parameter optimization is realized.

[0056] The application can accurately control the etching liquid flow rate, effectively improve the etching precision of the printed circuit board, and is especially suitable for differential etching control of precision lines, and significantly improves the product yield. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 The flow of the PCB precision line side etching suppression method based on vacuum etching is shown in the drawings Figure 1 ;

[0058] Figure 2is a flow of a PCB precision line side etching inhibition method based on vacuum etching Figure 2 . DETAILED DESCRIPTION

[0059] As shown in Figure 1-Figure 2 , the PCB precision line side etching inhibition method based on vacuum etching comprises the following steps:

[0060] Step S101, obtain the numerical distribution information of the line width, copper foil thickness and line density in the printed circuit board design file, extract the geometric feature parameters of each region through an image recognition algorithm, establish a multi-dimensional parameter matrix containing coordinate positions and feature values, and obtain a complete line feature data set;

[0061] Step S102, according to the line width W, copper foil thickness T and line density D in the multi-dimensional parameter matrix, a calculation model V=aW+bT+cD+d of etching liquid flow rate V is established by using a multivariate regression analysis algorithm, wherein a, b, c and d are regression coefficients, and the numerical values of each coefficient are determined by training historical etching data;

[0062] Step S103, if the line width is less than 0.1 millimeter and the line density is greater than a preset density threshold, a low flow rate control strategy is used to reduce the etching liquid flow rate to 70% of the reference value, and the etching time is extended to 1.3 times the reference time, so as to obtain the differential control parameters of the precision line region;

[0063] Step S104, generate the etching liquid flow rate distribution map corresponding to each region of the printed circuit board through the flow rate calculation model and the differential control parameters, supplement the flow rate transition values between adjacent regions by using an interpolation algorithm, and obtain a continuous flow rate control instruction sequence;

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

[0065] Step S106, if the deviation value exceeds a preset deviation threshold, a proportional-integral-derivative control algorithm is used to adjust the opening degree of the flow regulating valve, so that the actual flow rate converges to the target flow rate range, and it is judged whether the etching liquid temperature and concentration are within the standard range;

[0066] Step S107, real-time collection of line edge shape data in the etching process is realized by using an optical detection sensor, and the side etching degree value is calculated by using an edge detection algorithm, and if the side etching degree exceeds a preset side etching threshold, the flow rate parameters of the corresponding region are immediately adjusted;

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

[0068] As shown in Figure 1-Figure 2 Step S101, obtain the numerical distribution information of the line width, copper foil thickness, and line density in the printed circuit board design file, extract the geometric feature parameters of each region through an image recognition algorithm, establish a multi-dimensional parameter matrix containing coordinate positions and feature values, and obtain a complete line feature data set.

[0069] Further, in step S101, the original data is extracted from the printed circuit board design file through file parsing technology, and the line width, copper foil thickness, and line density are preliminarily classified to obtain structured numerical distribution information.

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

[0071] Through the region analysis method, the extracted geometric feature parameters are matched with the coordinate positions to construct a feature mapping table containing position information, and the correlation data of the features and coordinates are obtained.

[0072] According to the feature mapping table, the correlation between the coordinate positions and the feature values is analyzed, and a multi-dimensional parameter matrix containing the line width, copper foil thickness, and line density is generated using matrix construction technology to obtain a complete line feature set.

[0073] Through the multi-dimensional parameter matrix, data verification is performed on the distribution characteristics of the line width, copper foil thickness, and line density. If data is missing or abnormal, it is supplemented through interpolation method to obtain the corrected feature data set.

[0074] According to the corrected feature data set, the line feature set is classified and arranged in combination with the preset threshold value, it is judged whether the feature values of each region meet the design specification, and the line feature analysis result is obtained.

[0075] Through the analysis of the real-time collected data of the line width and line density, in combination with the line feature analysis result, the initial process condition judgment basis is obtained, and the process parameter adjustment direction is determined.

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

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

[0078] Specifically, in step S101, original data is extracted from a Gerber format or an ODB++ format of a printed circuit board design file through a file parsing technique, and a regular expression is used to match the numerical ranges of the line width, copper foil thickness and line density, for example, the line width extraction range is 0.05mm to 0.5mm, the copper foil thickness extraction range is 1oz to 3oz, and the line density is calculated by the total length of the line in a unit area. After preliminary classification, the structured JSON format numerical distribution information is generated;

[0079] According to the structured data, an edge detection algorithm in the OpenCV library is used to scan the graphic area in the design file, and the Hough transform is used to extract the straight line segment features, and the geometric feature parameters of each area are calculated, for example, the line width standard deviation is controlled within ±0.01mm, and the copper foil thickness area mean calculation accuracy reaches 0.1μm, and the feature description data is generated;

[0080] By 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 by using a quadtree spatial index, for example, the area with line width below 0.1mm is marked as high-precision area and is associated to the coordinates (x1, y1)-(x2, y2);

[0081] According to 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 1000x1000x3, and the complete line feature set is obtained;

[0082] Through data verification by the multi-dimensional parameter matrix, if it is found that the line width data of a certain area is missing, the bilinear interpolation method is used to supplement the missing value based on the data of the adjacent 4 coordinate points, and the copper foil thickness abnormal value is corrected by median filtering, to obtain the corrected feature data set.

[0083] According to the corrected data set, the line width tolerance threshold is preset to ±10%, the copper thickness tolerance is ±5%, and the density threshold is 80%, and the K-means clustering algorithm is used to classify the line features, to determine whether each area meets the design specification, and to output the classification label.

[0084] Through sliding window analysis of the real-time collected line width data stream, the window size is set to 50 sampling points, and combined with the line feature analysis result, if the average line width of the continuous 3 windows is less than 0.08mm and the density is more than 85%, it is determined that process adjustment is needed.

[0085] When the flow rate control strategy is triggered, the SVM model input parameters include the current flow rate 2.5 m / s and the etching time 120 s. The optimized output suggested flow rate 1.8 m / s and etching time 150 s are calculated by the kernel function. If the model output parameter deviation exceeds ±5%, the calculation is iterated again;

[0086] According to the optimized parameters, the etching process execution scheme is generated, and the specific conditions include that the flow rate is controlled at 1.8±0.1 m / s, the etching time is set to 150±5 s, and the temperature is maintained at 25±1℃.

[0087] As shown in Figure 1-Figure 2 step S102, a calculation model V=aW+bT+cD+d of the etching liquid flow rate V is established by using a multiple regression analysis algorithm according to the line width W, the copper foil thickness T, and the line density D in the multi-dimensional parameter matrix, where a, b, c, and d are regression coefficients, and the values of the coefficients are determined by training historical etching data.

[0088] Further, in step S102, an initial data set containing the line width W, the copper foil thickness T, the line density D, and the corresponding etching liquid flow rate V is obtained by historical etching data, and a basic data set for model training is constructed;

[0089] The initial data set is cleaned by using a data preprocessing method to eliminate missing values and abnormal values, and a standardized data set is obtained;

[0090] According to the standardized data set, a multiple regression analysis algorithm is applied to calculate the regression coefficients a, b, c, and d, and determine the calculation model V=aW+bT+cD+d of the etching liquid flow rate V;

[0091] The performance index data of the model is obtained by using a cross-validation method to evaluate the prediction accuracy of the calculation model;

[0092] If the model performance index is lower than the preset threshold, the feature weights in the data set are adjusted, the multiple regression analysis is re-executed, and the updated calculation model is obtained;

[0093] According to the updated calculation model, the data of the new line width W, the copper foil thickness T, and the line density D are input to predict the corresponding etching liquid flow rate V, and the prediction result value is obtained;

[0094] The applicability of the model in the production environment is verified by comparing the prediction result value with the actual etching liquid flow rate, and the final calculation model is determined;

[0095] The original data is extracted from the printed circuit board design file by using a file parsing technology, and the line width W, the copper foil thickness T, and the line density D required by the final calculation model are classified to obtain structured numerical distribution information;

[0096] According to the structured numerical distribution information, a multi-dimensional parameter matrix is constructed, and interpolation method is used to supplement the missing or abnormal data to obtain a corrected feature data set.

[0097] Specifically, in step S102, 100 groups of sample data are extracted from the historical etching database, including line width W (10-100 pm), copper foil thickness T (18-70 pm), line density D (20%-80%), and corresponding etching liquid flow rate V (1.5-5.2 m / s), to form an initial data set.

[0098] The data is cleaned by 3σ criterion, and the data points deviating from the mean value by ±3 times the standard deviation are removed, for example, the abnormal value of flow rate V>4.8 m / s, and 85 groups of effective data are reserved to form a standardized data set.

[0099] Based on the least square method, a multiple regression equation is fitted, the loss function is set as mean square error MSE, and the regression coefficients a=0.032, b=0.015, c=0.021, and d=1.207 are calculated by gradient descent iteration optimization.

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

[0101] The new parameters W=50 pm, T=35 pm, and D=45% are input, and the model outputs the predicted value V=3.21 m / s, with an error of 0.94% compared with the actual measured value 3.18 m / s on the production line.

[0102] The Gerber design file is analyzed, the line width area is recognized by OpenCV, the coordinates (X, Y) are extracted, W=48.7 pm, T=32.5 pm, D=43.6% is calculated by combining the pixel ratio of the area, and a 200x200 matrix is constructed.

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

[0104] As shown in Figure 1-Figure 2 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 used to reduce the etching liquid flow rate to 70% of the reference value, and the etching time is extended to 1.3 times the reference time, to obtain the differential control parameters for the precision line area.

[0105] Further, in step S103, by collecting the line width and line density data of the printed circuit board in real time, the initial process condition judgment basis is obtained by analysis;

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

[0107] According to the determined low flow rate value, the flow rate of the etching liquid is adjusted to 70% of the reference flow rate, and the adjusted flow rate parameter is obtained;

[0108] By adjusting the flow rate parameter, the etching time is extended by 1.3 times the reference time to obtain the differential control parameter of the precision line area;

[0109] The support vector machine model is used to optimize the differential control parameter, and whether the parameter deviation exceeds the preset range is judged;

[0110] If the optimization result shows that the parameter deviation exceeds the preset range, the flow rate value and the etching time are fine-tuned to obtain the final process control parameter;

[0111] According to the final process control parameter, the preliminary flow rate distribution of each area is generated combined with the flow rate calculation model, and the optimized regional flow rate distribution is determined;

[0112] The flow rate values between adjacent regions are smoothed by an interpolation algorithm, and if there is a mutation point, the interpolation algorithm parameter is adjusted, the flow rate value of the mutation region is recalculated, and a smooth flow rate distribution is obtained;

[0113] According to the smooth flow rate distribution, the corresponding flow rate control instruction sequence is generated, the running state of the etching liquid spraying equipment is adjusted, and the precise etching liquid flow rate output is obtained.

[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 laser range finder measuring a line width of 0.08 mm, and an image processing algorithm is used to calculate the number of lines per unit area. If the density reaches 5 lines per square millimeter and exceeds the preset threshold of 4 lines per square millimeter, the control strategy is triggered;

[0115] According to the preset rule, the etching liquid reference flow rate of 10 liters per minute is reduced to 7 liters per minute, and according to the time compensation formula, the reference etching time of 20 minutes is extended to 26 minutes;

[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 result shows that the flow rate deviation exceeds ±0.5 liters per minute or the time deviation exceeds ±1 minute, the parameters are automatically corrected to 6.8 liters per minute and 26.5 minutes;

[0117] The regional flow rate distribution map is generated by using a computational fluid dynamics model, and a cubic spline interpolation algorithm is used to process the transition between adjacent regions. When the flow rate difference between adjacent units exceeds 0.3 liters / minute, the interpolation node parameters are recalculated to make the transition gradient less than 0.1 liters / minute·mm;

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

[0119] As shown in Figure 1-Figure 2 Step S104, the flow rate calculation model and the differential control parameters are used to generate the etching liquid flow rate distribution map corresponding to each region of the printed circuit board. An interpolation algorithm is used to supplement the flow rate transition values between adjacent regions to obtain a continuous flow rate control instruction sequence.

[0120] Further, in step S104, the geometric feature data of each region of the printed circuit board and the initial flow rate parameters of the etching liquid are obtained, and the flow rate calculation model is input to generate preliminary flow rate distribution data;

[0121] According to the preliminary flow rate distribution data, the flow rate values of each region are adjusted in combination with the differential control parameters to obtain an optimized regional flow rate distribution;

[0122] An interpolation algorithm is used to smooth the flow rate values between adjacent regions in the optimized regional flow rate distribution to generate a continuous flow rate transition distribution;

[0123] If there is a flow rate discontinuity point in the continuous flow rate transition distribution, the flow rate values of the discontinuous region are recalculated by adjusting the interpolation algorithm parameters to obtain a smooth flow rate distribution;

[0124] According to the smooth flow rate distribution, a corresponding flow rate control instruction sequence is generated to determine the real-time flow rate control parameters of each region;

[0125] The real-time flow rate control parameters are used to adjust the operating state of the etching liquid injection equipment to obtain precise etching liquid flow rate output;

[0126] Real-time monitoring data of the etching liquid flow rate output is obtained and compared with the flow rate control instruction sequence to determine whether there is a deviation;

[0127] If the deviation between the monitoring data and the control instruction sequence exceeds a preset threshold, the flow rate calculation model parameters are adjusted to regenerate the flow rate distribution data;

[0128] According to the regenerated flow rate distribution data, 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 to obtain a flow rate output that meets the target.

[0129] Specifically, in step S104, by collecting the line width, spacing and copper thickness data of each region of the printed circuit board, combining the initial flow rate parameters such as the inlet pressure 2.5 Bar and the temperature 25℃, and using the finite element analysis method to establish a flow rate calculation model, the preliminary distribution data containing 5 sub-regions and the flow rate range of 1.2-3.8 m / s are output;

[0130] Based on the priority weight in the differentiated control parameter (high-precision area weight coefficient 1.5, ordinary area 1.0), the preliminary distribution is nonlinearly optimized, so that the flow rate of the key area is increased to 4.0 m / s;

[0131] The double cubic spline interpolation algorithm is used to process the adjacent sub-region boundary data to generate a transition distribution with a grid accuracy of 0.1 mm. If it is detected that the gradient change of a certain boundary exceeds 15%, local recalculation is triggered, the interpolation node density is adjusted to 0.05 mm, and the smooth curve is regenerated;

[0132] The final distribution is converted into a time-space matrix instruction sequence, including flow rate set values with an interval of 200 ms, and the opening degree of the electromagnetic valve (0-100% corresponding to 0.5-5.0 m / s) is adjusted through the PID controller;

[0133] Real-time monitoring is performed by using an ultrasonic flowmeter to sample (10 Hz frequency). When it is detected that the actual flow rate of B area 3.2 m / s deviates from the instruction value 3.5 m / s by more than 8%, the viscosity coefficient in the model is updated by the back propagation algorithm, a new distribution is iteratively generated, and the servo motor is driven to adjust the transverse movement speed of the nozzle array to 12 cm / s.

[0134] As shown in Figure 1-Figure 2 , in step S105, according to the flow rate control instruction 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 of the current flow rate from the target flow rate is obtained through the real-time monitoring algorithm.

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

[0136] Through the real-time monitoring algorithm, the current flow rate data of each nozzle is collected, the deviation value of the current flow rate from the target flow rate is calculated, and a flow rate deviation data set is obtained;

[0137] If the deviation value in the flow rate deviation data set exceeds the preset threshold value, the Kalman filtering algorithm is used to smooth the deviation data to obtain an optimized deviation correction value;

[0138] According to the optimized deviation correction value, a new flow rate control instruction sequence is generated, the control parameters of the time sequence and the position coordinates are updated, and a new valve driving scheme is determined;

[0139] Through the new valve driving scheme, the opening of the flow regulating valve is adjusted, the etching liquid flow rate output of each nozzle is controlled, and adjusted flow rate distribution data is obtained;

[0140] The real-time monitoring system is used to collect the adjusted flow rate distribution data, analyze the matching degree of the current flow rate and the target flow rate, and obtain a flow rate matching analysis result;

[0141] According to the flow rate matching analysis result, the instruction sequence database is updated, the parameters of the flow regulating model are optimized, and a continuously optimized flow rate control model is determined;

[0142] Through the optimized flow rate control model, combined 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 is a mutation point in the regional flow rate distribution, the flow rate value of the mutation region is recalculated by adjusting the interpolation algorithm parameters, and a smooth regional flow rate distribution is obtained.

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

[0145] Through the real-time monitoring algorithm, the nozzle flow rate data is collected with a sampling period of 10ms, the deviation of the current flow rate 48ml / s and the target value 50ml / s is calculated as 2ml / s, and a matrix data set containing the deviations of all nozzles is formed;

[0146] If there is a threshold item exceeding 1ml / s in the deviation matrix, the Kalman filter algorithm is used to set the process noise Q=0.1 and the observation noise R=1.0 for data smoothing, and the corrected deviation value 1.5ml / s is obtained;

[0147] According to the correction value, a new instruction sequence is generated, the control parameter at the coordinate (x1, y1) at the 100th ms is updated to the opening 65%, and a new driving scheme is formed;

[0148] After executing the new scheme, the adjusted flow rate data 49.8ml / s is collected by the electromagnetic flowmeter, and the matching degree with the target value 50ml / s reaches 99.6%;

[0149] The matching results are written into a SQL database, and the flow regulation model parameters are optimized using the gradient descent method with α = 0.01 to establish an updated control model.

[0150] Based on the geometric data of the PCB board with a length of 300 mm and a width of 200 mm, an initial distribution of flow rates of 45 ml / s in region A and 55 ml / s in region B is generated through the calculation model;

[0151] A flow rate jump of 10 ml / s is detected at the boundary of the regions, and a cubic spline interpolation algorithm is used to adjust the smoothing coefficient λ = 0.3 to recompute a continuous distribution of flow rates of 48 ml / s-52 ml / s in the transition region.

[0152] As shown in Figure 1-Figure 2 step S106, if the deviation value exceeds the preset deviation threshold, a proportional-integral-derivative control algorithm is used to adjust the flow regulation valve opening to make the actual flow rate converge to the target flow rate range, and it is determined whether the etching liquid temperature and concentration are within the standard range.

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

[0154] According to the adjusted valve opening, the flow regulation valve is controlled by the driving module to execute the opening change, and real-time flow rate data is obtained;

[0155] If the real-time flow rate data converges to the target flow rate range, the etching liquid temperature and concentration data are collected by the sensor to determine whether they meet the preset standard;

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

[0157] According to the predicted adjustment parameters, the heating or cooling device is adjusted by the temperature control module to obtain the adjusted etching liquid temperature;

[0158] According to the predicted adjustment parameters, the amount of chemical added is controlled by the concentration adjustment module to obtain the adjusted etching liquid concentration;

[0159] If the adjusted etching liquid temperature and concentration both meet the preset standard, the current flow rate temperature and concentration data are stored by the data recording module to determine that the process parameters are stable;

[0160] The side etching degree data and flow rate parameter information are obtained through the data acquisition system to obtain a structured side etching degree data set and a flow rate parameter data set;

[0161] The correlation between the two is calculated by the Pearson correlation coefficient method to determine the key influencing factors.

[0162] Specifically, in step S106, if the deviation of the actual flow rate from the target flow rate exceeds 5%, a 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] According to the adjusted valve opening, the flow regulating valve is controlled by the driving module to change the opening with a response time of 0.5 seconds, and real-time flow rate data (sampling frequency 10 Hz) are obtained;

[0164] If the real-time flow rate data converge to the target flow rate range (±2% error band), the etching liquid temperature (set value 25℃) and concentration (set value 10%) are collected by temperature sensors (accuracy ±0.1℃) and concentration sensors (accuracy ±0.5%), and it is determined whether they meet the preset standards (temperature allowable deviation ±1℃, concentration allowable deviation ±0.3%);

[0165] If the etching liquid temperature (26.5℃) or concentration (9.8%) deviates from the preset standard, a linear regression algorithm (historical data window 30 minutes, goodness of fit R 2 ≥0.95) is used to analyze the change trend of temperature and concentration, and the predicted adjustment parameters (temperature correction amount -1.2℃, concentration correction amount +0.2%) are obtained;

[0166] According to the predicted adjustment parameters, the heating or cooling device is adjusted by the temperature control module with PID algorithm (Kp=1.0, Ki=0.3, Kd=0.05), and the adjusted etching liquid temperature (25.3℃) is obtained;

[0167] According to the predicted adjustment parameters, the concentration adjusting module controls the amount of chemical added in a pulse addition manner (each addition amount 0.05L), and the adjusted etching liquid concentration (10.1%) is obtained;

[0168] If the adjusted etching liquid temperature and concentration both meet the preset standards, the current flow rate (12.5L / min), temperature (25.3℃) and concentration (10.1%) data are stored in CSV format by the data recording module, and it is determined that the process parameters are stable (continuous 3 times sampling meet the standards);

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

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

[0171] like Figure 1-Figure 2 As shown, in step S107, the line edge morphology data during the etching process is collected in real time by an optical detection sensor, and the side erosion degree value is calculated using an edge detection algorithm. If the side erosion degree exceeds a preset side erosion threshold, the flow rate parameters of the corresponding area are immediately adjusted.

[0172] Furthermore, in step S107, the line edge morphology data during the etching process is collected 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] Calculate the side erosion degree according to the edge profile data and generate a side erosion degree value;

[0175] If the side erosion degree exceeds the preset side erosion threshold, the area to be adjusted is determined and a regional adjustment signal is generated;

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

[0177] According to the adjusted flow rate data, new line edge morphology data is collected in real time to generate updated edge image data;

[0178] The Canny edge detection algorithm is used to process the updated edge image data, extract new edge contour features, and obtain the verified side erosion degree value;

[0179] If the verified lateral corrosion degree value still exceeds the preset lateral corrosion threshold, the lateral corrosion degree data and flow rate parameter information are recorded through the data acquisition system to generate a structured lateral corrosion degree data set and flow rate parameter data set;

[0180] Through correlation analysis, feature variables are extracted from the side etching degree dataset and the flow rate parameter dataset, and the correlation between the two is calculated using the Pearson correlation coefficient method to determine the key influencing factor.

[0181] Specifically, in step S107, the optical detection sensor collects a gray-scale image of the etched circuit edge at a rate of 10 frames per second, with an image resolution of 2048x1536 pixels, to generate 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-threshold processing (low threshold 50, high threshold 150) on the image to extract continuous edge pixel points and obtain edge profile data at sub-pixel level accuracy.

[0183] The side etching width is calculated based on the edge profile data, the least squares method is used to fit the ideal edge line, the actual edge deviation distance is measured, and if the side etching width exceeds the preset threshold of 5μm, the abnormal region coordinates (X1-Y1 to X2-Y2) are marked, and a region adjustment signal is generated;

[0184] The region adjustment signal triggers the flow rate controller to increase the etching liquid flow rate in the corresponding region from 1.2m / s to 1.5m / s, and generates an adjusted flow rate parameter table;

[0185] A new edge image is collected after adjustment, and the edge profile is re-extracted through the Canny algorithm, and the side etching width is calculated to be 4.8μm;

[0186] If the side etching width still exceeds the threshold, the current flow rate of 1.5m / s and the side etching width of 4.8μm are recorded to the SQL database to form a structured dataset containing a timestamp;

[0187] Correlation analysis is performed on 100 groups of historical data in the database, and features such as flow rate standard deviation and side etching mean are extracted, and the Pearson correlation coefficient r=0.92 is calculated to determine that flow rate fluctuation is the key influencing factor.

[0188] As shown in Figure 1-Figure 2 step S108, the correlation data of the side etching degree value and the flow rate parameter are obtained, the regression coefficients in the flow rate calculation model are updated through a machine learning algorithm, an adaptive parameter optimization mechanism is established, and continuously improved etching control precision is obtained.

[0189] Further, in step S108, the etching process is monitored in real time by a data acquisition system to obtain raw data of the side etching degree value and the flow rate parameter, and to obtain structured side etching degree dataset and flow rate parameter dataset;

[0190] According to the obtained side etching degree data set and the flow rate parameter data set, a correlation information analysis method is used to extract characteristic variables and determine key influence factors between the two;

[0191] The extracted characteristic variables are calculated by a Pearson correlation coefficient method, the correlation between the side etching degree value and the flow rate parameter is analyzed, and a correlation quantization result is obtained;

[0192] According to the correlation quantization result, a flow rate calculation model is constructed, a linear regression algorithm is used to initialize and train the model, and initial regression coefficients are obtained;

[0193] If the prediction error of the initial regression coefficients exceeds a preset threshold, the regression coefficients are iteratively optimized by a gradient descent method, and optimized regression coefficients are obtained;

[0194] The optimized regression coefficients are used to update the flow rate calculation model, and the model parameters are real-time corrected by an adaptive parameter adjustment mechanism, and an updated flow rate calculation model is obtained;

[0195] The etching process is real-time controlled by the updated flow rate calculation model, etching control precision data is collected, and it is judged whether the precision meets the preset threshold requirement;

[0196] If the etching control precision does not meet the preset threshold requirement, the correlation information analysis step is returned, the characteristic variables are re-extracted and the model parameters are optimized, and an improved flow rate calculation model is obtained;

[0197] According to the improved flow rate calculation model, a new control instruction is generated, the etching process parameters are adjusted combined with real-time monitoring data, and continuously optimized etching control precision data is determined.

[0198] Specifically, in step S108, the data acquisition system is used to real-time monitor the side etching depth and the corresponding nozzle flow rate of 12 regions in the etching tank at a sampling frequency of 100Hz, the collection period is 5 seconds, and a structured CSV data set containing time stamp, position coordinates, side etching micron level measurement value and flow rate L / min data is generated;

[0199] The principal component analysis method is used to extract the first three characteristic components with a contribution rate of more than 85% from the data set as key influence factors, including peak etching rate, average flow rate gradient and transient flow pressure fluctuation coefficient;

[0200] The Pearson correlation coefficient is used to calculate that the correlation coefficient of the side etching depth and the peak flow rate is 0.78, and 4 strong correlation variables are selected when the significance level p<0.01 is set;

[0201] Based on the screening variables, a multiple linear regression model y = β0+ β1x1+... + β4x4 was constructed. The initial coefficients β0= 2.35, β1= 0.63 were obtained by least square method, and the mean square error (MSE) was 4.2 μm;

[0202] When the MSE exceeds the preset threshold 3.0 μm, the random gradient descent method with learning rate η = 0.01 is used for 200 iterations to make the regression coefficients converge to β0' = 2.18, 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 last 100 groups of data for re-fitting, and the variable weight of the β coefficient deviation exceeding 5% is 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 ± 0.5 L / min range. The laser speedometer feeds the flow rate data to the control end in real time;

[0205] If the standard deviation of the uniformity of the continuous 3 times of adjustment is still > 1.2 μm, the feature re-extraction process is triggered, and the kernel density estimation method is used to re-divide the flow rate-etching amount nonlinear relationship area;

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

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

Claims

1. A method for suppressing lateral corrosion of PCB precision circuits based on vacuum etching, characterized in that: include: Obtain circuit feature data from the printed circuit board design file and construct a multi-dimensional parameter matrix containing coordinate positions and feature values; Establishing a calculation model for the flow rate of the etching solution using a regression analysis algorithm based on the circuit characteristic variables in the multidimensional parameter matrix; Generate the flow velocity distribution of each area through the calculation model, and adjust the flow velocity value in combination with the differentiated control parameters; Generate a flow rate control instruction sequence according to the adjusted flow rate value, and drive the etching equipment to adjust the etching liquid flow rate of each nozzle; Obtain the deviation data between the current flow rate and the target flow rate through real-time monitoring, and adjust the flow control valve to make the flow rate converge to the target range; Collect line edge morphology data during etching, calculate the degree of side erosion and update the flow rate parameters based on the results; The regression coefficients in the calculation model are optimized by correlation data, and an adaptive parameter optimization mechanism is established.

2. A method for suppressing lateral corrosion of PCB precision circuits based on vacuum etching according to claim 1, characterized in that: The multidimensional parameter matrix includes: Extract the numerical distribution information of line width, copper foil thickness and line density from the Gerber format or ODB++ format of the printed circuit board design file; The geometric feature parameters of each region are extracted through edge detection algorithm and Hough transform; 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 1000×1000×3; When there is missing data, bilinear interpolation is used to supplement the missing values ​​based on the data of adjacent coordinate points.

3. The method for suppressing lateral corrosion of PCB precision circuits based on vacuum etching according to claim 1, characterized in that: The adjusting flow rate value includes: When the line width is less than 0.1 mm and the line density is greater than the preset density threshold, the low flow rate control strategy is triggered; Reduce the etching solution flow rate to 70% of the baseline value and extend the etching time to 1.3 times the baseline time; The flow rate and etching time are optimized by the support vector machine model, and the iterative calculation is repeated when the output parameter deviation exceeds ±5%.

4. The method for suppressing lateral corrosion of PCB precision circuits based on vacuum etching according to claim 1, characterized in that: The flow rate control instruction sequence includes: Generate preliminary velocity distribution maps for each area based on the velocity calculation model and differentiated control parameters; The bicubic spline interpolation algorithm is used to supplement the velocity transition values ​​between adjacent areas with a grid accuracy of 0.1 mm; When it is detected that the velocity gradient change in adjacent areas exceeds 15%, the interpolation node density is adjusted to 0.05 mm and recalculated; The final flow velocity distribution is converted into an instruction sequence matrix containing timestamps and spatial coordinates with a time interval of 200ms.

5. The method for suppressing lateral corrosion of PCB precision circuits based on vacuum etching according to claim 1, characterized in that: The adjusting the flow regulating valve to converge the flow rate to the target range includes: When the flow rate deviation exceeds 5%, the proportional-integral-differential control algorithm is used to adjust the valve opening, where the proportional coefficient Kp=0.8, the integral coefficient Ki=0.2, and the differential coefficient Kd=0.

1.

6. The method for suppressing lateral corrosion of PCB precision circuits based on vacuum etching according to claim 1, characterized in that: The step of calculating the lateral erosion degree and updating the flow velocity parameters according to the result includes: An optical detection sensor is used to capture grayscale images of the line edge at 10 frames per second, and the Canny edge detection algorithm is used to calculate the side erosion width. If the side erosion width exceeds 5μm, the flow rate in that area is increased by 0.3m / s.

Images