A fully automatic testing equipment for PCB boards

Abstract

The invention provides a fully automatic testing device for a PCB board, relates to the technical field of automation equipment, and comprises a detection unit, a transmission unit and a control unit; the detection unit comprises: a first image collector, a second image collector, an online X-ray scanner and a labeling machine, which are used for detecting and tracking the PCB; the transmission unit comprises: a first conveyor belt, a second conveyor belt and a PCB flipping piece, which are used for conveying and flipping the PCB; the control unit comprises: a data acquisition module, a data processing module, an action adjustment module and a defect positioning module, which are used for collecting and processing PCB image data, adjusting the running speed of the transmission unit and tracking the data of the PCB; the invention completes the flipping of the PCB during transmission by using the PCB flipping piece, so as to improve the testing efficiency of the PCB board, and the testing speed of the PCB board can be adjusted according to production requirements by using the control unit.

Description

Technical Field

[0001] The present invention relates to the technical field of automation equipment, in particular to a fully automatic testing equipment for PCB boards. Background Art

[0002] As the core conductive component of electronic products, the quality of bare printed circuit boards (PCBs) directly affects the overall performance. Due to the complexity of the manufacturing process, various defects will inevitably occur when producing bare PCB products. Therefore, it is necessary to test the PCB boards after production to avoid soldering components to defective bare PCBs in the next step, which would cause greater losses.

[0003] For large-scale assembly-line PCB board testing, existing PCB board testing equipment usually needs to keep the PCB board fixed during the transmission process to facilitate image acquisition by the visual acquisition device, and needs to be equipped with a fixture suitable for the PCB board or use a mechanical arm to turn the PCB board over to complete the full-surface image acquisition of the PCB board;

[0004] The use of fixtures limits the adaptability of test equipment to PCB board specifications and shapes. The movement efficiency of the robotic arm makes it easy to limit the testing efficiency of PCB boards, thus creating a bottleneck in the assembly line, which makes it difficult to improve the testing efficiency of the equipment. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a fully automatic testing device for PCB boards. By using a PCB turning piece, the PCB can be turned over during transportation to improve the testing efficiency of the PCB boards. By using a control unit, the testing speed of the PCB boards can be adjusted according to production requirements.

[0006] In order to solve the above technical problems, the technical solution of the present invention is as follows:

[0007] A fully automatic testing device for a PCB board, comprising a detection unit, a transmission unit and a control unit;

[0008] The detection unit comprises:

[0009] The first image collector is used to collect the front image information of the PCB board;

[0010] The second image collector is used to collect image information of the back side of the PCB board;

[0011] Online X-ray scanner, used to collect internal image information of PCB boards;

[0012] Labeling machine, used to track and record PCB boards;

[0013] The transmission unit comprises:

[0014] A first conveyor belt is used to convey the PCB board so that the first image collector completes the acquisition action;

[0015] A second conveyor belt is used to convey the PCB board so that the second image collector and the online X-ray scanner can complete the collection action;

[0016] A PCB flipping member, used for flipping the PCB board so as to complete the flipping action when the PCB board is transferred from the first conveyor belt to the second conveyor belt;

[0017] The PCB flipping member comprises a plurality of flipping boxes arranged in a circular array between the first conveyor belt and the second conveyor belt, the interiors of the flipping boxes are all slidably connected with push blocks, the surface of the first conveyor belt is equidistantly fixedly connected with a plurality of push plates, the PCB board on the surface of the first conveyor belt is pushed into the interior of the push blocks by the push plates, and the PCB board is transferred from the surface of the first conveyor belt to the surface of the second conveyor belt after being flipped by the push blocks through the linkage between the corresponding two push blocks;

[0018] The control unit comprises:

[0019] A data acquisition module, used to process image information and collect and transmit the operation data of the machine tool;

[0020] The data processing module eliminates the placement deviation through image alignment according to the image information, and finally outputs the image data, which is used to input the image data into the trained first learning model to identify the defect data of the PCB board and output the defect level of the PCB board;

[0021] An action adjustment module, used to screen the PCB for waste products at each image acquisition stage of the detection unit according to the defect level, obtain a waste skip rate, so as to reduce the excessive use of the PCB flipping parts, and dynamically adjust the speeds of the first conveyor belt and the second conveyor belt according to the real-time change of the waste skip rate, so as to control the comprehensive cost in the equipment detection process;

[0022] The defect location module is used to generate a defect location map based on the defect data of the PCB board.

[0023] Furthermore, the trained first learning model is obtained by the following steps:

[0024] Collect several groups of historical test data as sample sets, and perform data enhancement on the sample sets to obtain enhanced sample sets;

[0025] The enhanced sample set is randomly divided into a training set, a validation set and a test set. The image data in the training set is used as the input of the first learning model, and the defect level in the training set is used as the output of the first learning model. The first learning model is trained, and the training process of the first learning model is monitored through the validation set. The performance of the first learning model is evaluated through the test set, and finally the first learning model that meets the preset accuracy is output.

[0026] Furthermore, the PCB is screened for scrap according to the defect level at each image acquisition stage of the inspection unit to obtain the scrap skip rate to reduce the excessive use of PCB flip parts, including:

[0027] Score the PCB at each acquisition stage, set a dynamic screening threshold, and determine the acquisition action for the next stage based on the dynamic screening threshold;

[0028] Filter actions include:

[0029] In the process of collecting graphic data, the defect data in the image data is identified through the trained first learning model, and the corresponding defect level is output. If the PCB is marked as scrap, the subsequent stages are skipped.

[0030] Furthermore, the speeds of the first conveyor belt and the second conveyor belt are dynamically adjusted according to the real-time change of the scrap skip rate to control the comprehensive cost in the equipment detection process. The calculation formula of the comprehensive cost is:

[0031] ;

[0032] in, and are the speed power consumption coefficients of the first conveyor belt and the second conveyor belt, and are the speeds of the first conveyor belt and the second conveyor belt, and are the basic power consumption of the first conveyor belt and the second conveyor belt respectively, and Friction loss caused by the operation of the first conveyor belt and the second conveyor belt; and is the weight coefficient; is the length of a single PCB board, is the number of PCB boards to be tested in the equipment, is the minimum speed of the two conveyor belts.

[0033] Furthermore, a defect location map is generated based on the defect data of the PCB board, including:

[0034] After detection, an image marking the defect position is generated, the generated image is stored as a file and associated with the PCB number, and the barcode attached to the corresponding PCB board by the labeling machine is scanned by a scanning device, so that the terminal device displays the corresponding defect location map.

[0035] Furthermore, the PCB flip member further includes:

[0036] A column is arranged between the first conveyor belt and the second conveyor belt, and a rotating flower rod is symmetrically arranged between the two columns. The pushing block, the connecting block and the push plate are elastic. The pushing blocks are fixedly connected to the two rotating flower rods. A pushing block is installed on the surface of one of the columns, and the output shaft of the pushing block is fixedly connected to the corresponding rotating flower rod.

[0037] Furthermore, a plurality of connecting blocks are symmetrically arranged between the two corresponding pushing blocks, and the connecting blocks are fixedly connected to the two corresponding pushing blocks. The length of the connecting blocks is the sum of the length of the flip box along the extension direction of the first conveyor belt and the distance between the two corresponding flip boxes.

[0038] Furthermore, a first arc-shaped guard plate and a second arc-shaped guard plate are arranged between the two rotating flower rods, and the first arc-shaped guard plate and the second arc-shaped guard plate are both fixedly connected to the first conveyor belt and the second conveyor belt.

[0039] Furthermore, a plurality of first rotating rollers and a plurality of second rotating rollers are arranged inside the flip box, the first rotating rollers are rotatably connected to the flip box via two first one-way bearings, the second rotating rollers are rotatably connected to the flip box via two second bearings, the rotation directions of the first rotating roller and the second rotating roller are opposite, and the first rotating roller and the second rotating roller are in conflict with the corresponding pushing block and the connecting block.

[0040] Furthermore, the surfaces of the first rotating roller and the second rotating roller are not smooth and are elastic, the inner walls of the first arc-shaped guard plate and the second arc-shaped guard plate are smooth and elastic, and the pushing block and the pushing plate are elastic.

[0041] The above solution of the present invention includes at least the following beneficial effects:

[0042] The above solution of the present invention, through the use of the PCB flipping member, can complete the flipping action of the PCB board during the process of collecting image information on the conveyor belt, without the auxiliary use of a clamp or a mechanical arm, thereby reducing the test cost of the PCB;

[0043] Through the setting of the control unit, the test speed of the PCB can be adjusted according to the scrap skip rate and the comprehensive cost to meet different production needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a schematic diagram of the overall structure provided by the present invention.

[0045] Figure 2 It is a schematic diagram of the rotating flower rod in the present invention.

[0046] Figure 3 It is a schematic diagram of the flip box in the present invention.

[0047] Figure 4 It is a schematic diagram of the connection block in the present invention.

[0048] Figure 5 It is a schematic diagram of the first curved plate in the present invention.

[0049] Figure 6 The present invention Figure 5 Enlarged view of point A in the middle.

[0050] Figure 7 The present invention Figure 5 Enlarged view of point B in the middle.

[0051] Figure 8 The present invention Figure 5 Enlarged view of point C in the middle.

[0052] Fig. 9 The present invention Figure 5 Enlarged view of point D in the middle.

[0053] Fig.10 It is a schematic diagram of the control unit in the present invention.

[0054] In the figure: 101, first conveyor belt; 102, second conveyor belt; 103, PCB turning piece;

[0055] 201, first image collector; 202, second image collector; 203, online X-ray scanner; 204, labeling machine;

[0056] 301, column; 302, rotating flower rod; 303, flip box; 304, push block; 305, connecting block; 306, first rotating roller; 307, second rotating roller; 308, first arc-shaped guard plate; 309, second arc-shaped guard plate; 310, push plate. DETAILED DESCRIPTION

[0057] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0058] like Figures 1 to 10 As shown, an embodiment of the present invention provides a fully automatic testing device for a PCB board, including a detection unit, a transmission unit and a control unit;

[0059] The detection unit includes:

[0060] The first image collector 201 is used to collect the front image information of the PCB board;

[0061] The second image collector 202 is used to collect image information of the back side of the PCB board;

[0062] An online X-ray scanner 203 is used to collect internal image information of the PCB board;

[0063] A labeling machine 204 is used to track and record the PCB boards;

[0064] The transmission unit includes:

[0065] A first conveyor belt (101) is used to convey the PCB board so that the first image collector 201 completes the acquisition action;

[0066] The second conveyor belt 102 is used to convey the PCB board so that the second image collector 202 and the online X-ray scanner 203 can complete the collection action;

[0067] The PCB flipping member 103 is used to flip the PCB board so as to complete the flipping action when the PCB board is transferred from the first conveyor belt 101 to the second conveyor belt 102;

[0068] The PCB flipping member 103 includes a plurality of flipping boxes 303 arranged in a circular array between the first conveyor belt 101 and the second conveyor belt 102. The interior of the flipping boxes 303 is slidably connected with a push block 304. The surface of the first conveyor belt 101 is fixedly connected with a plurality of push plates 310 at equal intervals. The push plates 310 are used to push the PCB board on the surface of the first conveyor belt 101 into the interior of the push block 304. Through the linkage between the two corresponding push blocks 304, the PCB board is transferred from the surface of the first conveyor belt 101 to the surface of the second conveyor belt 102 after being flipped by the push block 304.

[0069] The control unit includes:

[0070] A data acquisition module, used to process image information and collect and transmit the operation data of the machine tool;

[0071] A data processing module, used to eliminate placement deviations through image alignment according to image information, and finally output image data, input the image data into the trained first learning model to identify defect data of the PCB board, and output the defect level of the PCB board;

[0072] The action adjustment module is used to screen the PCB for waste products according to the defect level in each image acquisition stage of the detection unit to obtain the waste skip rate, so as to reduce the excessive use of the PCB flipping member 103, and dynamically adjust the speed of the first conveyor belt 101 and the second conveyor belt 102 according to the real-time change of the waste skip rate, so as to control the comprehensive cost in the equipment detection process;

[0073] The defect location module is used to generate defect location according to the defect data of the PCB board.

[0074] In the embodiment of the present invention, the first image collector 201 and the second image collector 202 are equipped with high-definition cameras or optical sensors to capture clear images of the front of the PCB board. The online X-ray scanner 203 collects internal image information of the PCB board through X-ray imaging technology to facilitate the collection of the data acquisition module. The labeling machine 204 labels each PCB board in the form of a barcode or a QR code to track its position and status during the test process, thereby ensuring the traceability of the production and test process of each PCB board and avoiding confusion.

[0075] The PCB board to be tested can be placed on the conveyor belt of the first conveyor belt 101 by using a PCB placing machine, so that the first image collector 201 completes the acquisition action. The PCB flipping piece 103 is used to flip the PCB board, so that the flipping action is completed when the PCB board is transferred from the first conveyor belt 101 to the second conveyor belt 102, so that the second image collector 202 and the online X-ray scanner 203 complete the acquisition action. Specifically, when the PCB board on the surface of the first conveyor belt 101 is conveyed to the left end position of the first conveyor belt 101, the PCB board begins to enter the flip box 303 close to the first conveyor belt 101. During this process, the PCB will slide relative to the conveyor belt surface of the first conveyor belt 101. The PCB board is pushed into the flip box 303 by the adjacent push plate 310. During this process, the PCB board pushes the push block 304 to move in the flip box 303 toward the second conveyor belt 102 until the PCB board completely enters the flip box 303. If the other flip box 303 corresponding to the flip box 303 contains a PCB board, the push blocks 304 in the two flip boxes 303 are linked to each other, so that during the process of the PCB on the first conveyor belt 101 entering the flip box 303, the PCB board in the other flip box 303 corresponding to the flip box 303 is pushed out, so that the pushed out PCB board falls on the surface of the second conveyor belt 102.

[0076] The data acquisition module processes the image provided by the detection unit, corrects the deviation in the image, and finally generates image data. Through image alignment and deviation correction, the accuracy of the collected image data is ensured to avoid erroneous judgments caused by improper PCB board position;

[0077] The data processing module analyzes the image data through the first learning model, identifies the defects on the PCB board, and outputs the defect level. Specifically, the defect level can be divided into qualified level, repair level and scrap level;

[0078] The action adjustment module adjusts the detection action and the running speeds of the first conveyor belt 101 and the second conveyor belt 102 according to the defect level, so that the detection action skips the waste-level PCB boards. When waste-level PCB boards appear at the same time, the running speeds of the second conveyor belt 102 and the first conveyor belt 101 are adjusted to improve the overall detection efficiency of the PCB boards;

[0079] The defect location module generates an accurate defect location map based on the defect data of the PCB board, showing the specific location and nature of the defect, providing detailed defect information for maintenance and quality control, and helping technicians to quickly locate and repair.

[0080] like Figures 1 to 10 As shown, the training method of the first learning model includes:

[0081] Several groups of historical test data are randomly collected as sample sets, and data enhancement is performed on the sample sets. The enhanced sample sets are randomly divided into training sets, validation sets and test sets. The image data in the training set is used as the input of the first learning model, and the defect level in the training set is used as the output of the first learning model. The first learning model is trained, the training process of the first learning model is monitored through the validation set, the performance of the first learning model is evaluated through the test set, and finally the first learning model that meets the preset accuracy is output.

[0082] In an embodiment of the present invention, several groups of PCB board images after testing in the production history are collected, and a labeling tool such as LabelImg or Labelme can be used to label the position of each defect on each PCB board with a bounding box or a pixel-level mask to generate a labeling file, and a defect level labeling is performed on each PCB board in each of the several groups of data. The defect level includes a qualified level, a repair level, and a scrap level. The labeling of the defect level can be performed by manual labeling. By performing set transformations on the image, such as rotation, translation, and mirroring, color transformations, such as adjusting brightness, contrast, and saturation, and adding noise, the interference in actual acquisition is simulated, the adaptability of the first learning model to different scenarios is improved, overfitting is prevented, and the robustness of the first learning model is improved;

[0083] Qualified PCBs have no obvious defects, or only minor defects that will not affect the performance of the PCB; Repaired PCBs have obvious defects that can be repaired to not affect the function of the PCB; Scrap PCBs have defects that cannot be repaired and affect the function of the PCB.

[0084] Set training objectives: Identify defective areas in PCB images by training the first learning model, use regression loss functions, such as bounding box regression, and output the bounding box of each defective area;

[0085] The first learning model is trained using the cross entropy loss function to output the defect level of the entire PCB board;

[0086] You can use convolutional neural networks (CNNs) and object detection frameworks such as Faster R-CNN, YOLO, or RetinaNet, which can handle both object detection and classification tasks;

[0087] Several groups of historical test data are divided into training set, validation set and test set in a ratio of 7:2:1. The training set is the main data set used by the model for actual learning. The model uses the data in the training set to adjust its parameters during the training process. The validation set is used to monitor the model training process and evaluate the performance of the model on data outside the training. It does not participate in parameter optimization. The test set is used to evaluate the final performance of the model after the model training is completed. The test results reflect the model's generalization ability to unseen data.

[0088] Training architecture:

[0089] Convolutional layer (CNN): extracts features from the image.

[0090] Region Proposal Network (RPN): Generates candidate regions, i.e., regions that may contain defects.

[0091] Bounding box regression: Accurately locate each candidate region.

[0092] Classification head: classifies the defect area and determines whether the area belongs to a certain type of defect.

[0093] Overall classification: Classify the entire PCB and determine whether it is qualified, repaired or scrap.

[0094] The loss function is , specifically: ;

[0095] Loss Function middle, is the classification loss, which is used to optimize the classification of defect areas;

[0096] ;

[0097] in, is the predicted probability, Make the real label;

[0098] Loss Function middle, is the regression loss, which is used to optimize the border position of the defect area;

[0099] ;

[0100] in, is the bounding box predicted by the first learned model, is the ground-truth bounding box, Indicated as the first samples;

[0101] Loss Function middle, It is an overall loss classification, which is used to classify defects in the entire PCB image; among them:

[0102] ;

[0103] in, is the true label of the overall PCB classification, is the probability predicted by the first learned model, Indicated as the first samples, is a coefficient used to balance the target detection loss and the overall classification loss, which can be adjusted by cross-validation.

[0104] Training process:

[0105] Input the PCB image, the first learning model extracts features through the convolution layer, the region proposal network (RPN) generates candidate regions, and classifies and regresses each region, and finally classifies the entire PCB image; according to the predicted category and bounding box position output by the model, the loss function of each sample is calculated, and the loss function is back-propagated to adjust the network weights to minimize the loss; optimization algorithms such as Adam or SGD are usually used for optimization and weight update.

[0106] Model Evaluation:

[0107] mAP (Mean Average Precision) is often used as an evaluation indicator for target detection. The average accuracy of all detection categories is calculated. The accuracy and recall of each detection box are evaluated through Precision and Recall. The accuracy of the model in classifying the overall defect level of the PCB is evaluated through the accuracy rate. The F1 score is a weighted average of Precision and Recall to balance the performance of the evaluation model.

[0108] Result output: Output the bounding box and category label of each detected defect, output the defect level of each PCB detection, and take the worst level (scrap, worse than repair, worse than qualified) as the defect level of the entire PCB, which is the classification label of the PCB board.

[0109] like Figure 1 and Fig.10 As shown, the PCB is screened for waste products at each image acquisition stage of the detection unit according to the defect level, and the waste skip rate is obtained to reduce the excessive use of the PCB flipping member 103, including:

[0110] Score the PCB at each acquisition stage, set a dynamic screening threshold, and determine the acquisition action for the next stage based on the dynamic screening threshold;

[0111] Filter actions include:

[0112] In the process of collecting graphic data, the defect data in the image data is identified through the trained first learning model, and the corresponding defect level is output. If the PCB is marked as scrap, the subsequent stages are skipped.

[0113] In the embodiment of the present invention, during the process of testing the PCB board, the image information of the PCB board is collected by the first image collector 201, the second image collector 202 and the online X-ray scanner 203, and the collected image data is subjected to denoising, graying and standardization operations, and then passed to the trained first learning model as input, and all possible defect areas are identified in the input image, and each defect area is classified and its category is identified. Based on the type and severity of the defect, the first learning model outputs the defect level of each test of the PCB board. If a PCB board of waste grade is evaluated at any stage of the first image collector 201 and the second image collector 202, the subsequent test of the PCB board is skipped, thereby saving time and resources and increasing the waste skipping rate. The percentage of skipped scraps to the total number of tested PCBs.

[0114] In order to decide whether to continue collecting data according to different situations, a dynamic screening threshold needs to be set. This threshold is based on the comprehensive consideration of historical data, defect distribution in the production process, and real-time collection data;

[0115] According to production needs, if the quality of the PCB board is high, the threshold will be higher so that smaller defects can be included in the repair range to ensure that the PCB board meets the requirements after repair; if certain defect types often lead to the appearance of waste PCB boards, the threshold will be lowered to facilitate the screening of waste in the early stage, thereby skipping the inspection of subsequent stages in advance. The dynamic screening threshold determines whether the model needs to continue to collect image data at each stage and whether to enter the next test stage.

[0116] If the first learning model outputs a waste grade, it is determined that the PCB does not need to continue with the subsequent test steps (if any), otherwise the subsequent image acquisition (if any) is continued, thereby saving the resources of the test equipment, reducing energy consumption, and improving the overall efficiency, wherein the collection action of the online X-ray scanner 203 is reduced, and the energy consumption is reduced more;

[0117] like Figure 1 and Fig.10 As shown, the speeds of the first conveyor belt 101 and the second conveyor belt 102 are dynamically adjusted according to the real-time change of the scrap skip rate to control the comprehensive cost during the equipment detection process. The calculation formula of the comprehensive cost is:

[0118] ;

[0119] in, and are the speed power consumption coefficients of the first conveyor belt 101 and the second conveyor belt 102, respectively, and are the speeds of the first conveyor belt 101 and the second conveyor belt 102, respectively, and are the basic power consumptions of the first conveyor belt 101 and the second conveyor belt 102 respectively, and The friction loss of the first conveyor belt 101 and the second conveyor belt 102; and is the weight coefficient; is the length of a single PCB board, is the number of PCB boards to be tested in the equipment, is the minimum speed of the two conveyor belts.

[0120] In an embodiment of the present invention, dynamically adjusting the speeds of the first conveyor belt 101 and the second conveyor belt 102 according to the real-time change of the scrap skip rate includes:

[0121] pass Calculate the dynamic speed of the first conveyor belt 101 by The dynamic speed of the second conveyor belt 102 is calculated, where is the dynamic speed of the first conveyor belt 101, is the dynamic speed of the second conveyor belt 102, and is the adjustment coefficient, and are the basic speeds of the first conveyor belt 101 and the second conveyor belt 102, respectively, is the scrap skip rate.

[0122] and The first conveyor belt 101 and the second conveyor belt 102 represent the default conveyor belt speeds when there is no impact of the scrap skip rate. and It is an adjustment factor used to control the influence of scrap skip rate on speed. Usually, and The value can be adjusted according to actual needs to adapt to different production environments and quality requirements, such as > It can be indicated that when the scrap skipping rate is high, the speed of the first conveyor belt 101 tends to be increased.

[0123] and Indicates that the speed of the conveyor belt is adjusted by the scrap skip rate. When the scrap skip rate is high, the speed of the first conveyor belt 101 will be relatively faster, and the speed of the second conveyor belt 102 will be maintained or appropriately increased to reduce unnecessary subsequent tests. When the scrap skip rate is low, the speed of the first conveyor belt 101 will slow down, and the speed of the second conveyor belt 102 It will also be slowed down appropriately to ensure that all PCB boards are fully tested.

[0124] Adjustment and The goal is to balance test efficiency, resource consumption, and energy usage.

[0125] For example, in the scenario where the short-term impact of the scrap skip rate on PCB board detection is considered, after the scrap is detected at the first image collector 201, a real-time When the waste leaves the detection area of ​​the online X-ray scanner 203, the real-time The value will no longer take into account the influence of the waste. In this way, the residence time of the waste in the detection area of ​​the first image collector 201, the second image collector 202 and the online X-ray scanner 203 can be shortened. A robotic arm can also be set between the first image collector 201 and the second image collector (202), and between the second image collector 202 and the online X-ray scanner 203. After the waste is detected, the waste can be removed in time to increase the speed adjustment range of the first conveyor belt 101 and the second conveyor belt 102.

[0126] When production demand tends to control the comprehensive cost of equipment, it can be achieved through Calculate the comprehensive cost value, where and are the speed power consumption coefficients of the first conveyor belt 101 and the second conveyor belt 102, respectively, and are the basic power consumptions of the first conveyor belt 101 and the second conveyor belt 102 respectively, and is the friction loss of the first conveyor belt 101 and the second conveyor belt 102, and is the weight coefficient, is the length of a single PCB board, is the number of PCB boards to be tested in the equipment, is the minimum speed of the two conveyor belts.

[0127] In the calculation of comprehensive cost, represents the energy consumption of the first conveyor belt 101 and the second conveyor belt 102 during operation, wherein, and The basic power consumption represented may be the energy consumption of the first conveyor belt 101 and the second conveyor belt 102 in a stationary state. Used to balance the relative importance of different types of energy consumption; specifically, for the friction loss power consumption of the second conveyor belt 102; wherein:

[0128] ;

[0129] in, is the friction coefficient of the first conveyor belt 101, is the weight of the PCB on the conveyor belt, is the acceleration due to gravity, and the friction loss power consumption of the second conveyor belt 102; wherein:

[0130] ;

[0131] in, is the friction coefficient of the second conveyor belt 102;

[0132] In the calculation of comprehensive cost, This part takes into account the impact of production efficiency on cost. It reflects that the faster the conveyor belt speed is, the more PCB boards can be processed per unit time, thus reducing the processing time cost per unit product; if the comprehensive cost increases, the cost can be reduced by slowing down the speed, and if the comprehensive cost decreases, the efficiency can be improved by speeding up the speed;

[0133] like Figure 1 and Fig.10 As shown, a defect location map is generated based on the defect data of the PCB board, including:

[0134] After the detection, an image marking the defect position is generated, the generated image is stored as a file and associated with the PCB number, and the barcode attached to the corresponding PCB board by the labeling machine 204 is scanned by a scanning device, so that the terminal device displays the corresponding defect location map.

[0135] In an embodiment of the present invention, based on the defect recognition result of the first learning model, the defect position is marked in the image, and a rectangular frame is usually drawn to mark the specific defect area. If there are multiple defects, multiple marking frames will be displayed in the image at the same time. The processed image is saved as an image file, such as JPEG or PNG format. The generated image file includes all detected defect position marks, which can be viewed and printed in subsequent analysis. The labeling machine 204 is used to label each PCB board model passing through the online X-ray scanner 203, and each PCB board is numbered. The number corresponds to the barcode affixed by the labeling machine 204. The number is tracked throughout the entire production process, and the generated defect location map is associated with the PCB number to ensure that each image file corresponds to the correct PCB board, and the image data is saved in a database or server so that the corresponding defect location map can be found according to the PCB number.

[0136] When all PCB boards have been inspected, they can be classified by scanning equipment. For PCB boards that need to be repaired, the terminal equipment, such as a computer, will display the corresponding defect location map for the operator to view.

[0137] like Figures 2 to 9 As shown, the PCB flip member 103 also includes:

[0138] A column 301 is arranged between the first conveyor belt 101 and the second conveyor belt 102, and a rotating flower rod 302 is symmetrically arranged between the two columns 301. The pushing block 304, the connecting block 305 and the push plate 310 are elastic. The pushing blocks 304 are fixedly connected to the two rotating flower rods 302. A pushing block 304 is installed on the surface of one of the columns 301, and the output shaft of the pushing block 304 is fixedly connected to the corresponding rotating flower rod 302.

[0139] A number of connecting blocks 305 are symmetrically arranged between the two corresponding pushing blocks 304. The connecting blocks 305 are fixedly connected to the two corresponding pushing blocks 304. The length of the connecting blocks 305 is the sum of the length of the flip box 303 along the extension direction of the first conveyor belt 101 and the distance between the two corresponding flip boxes 303.

[0140] A first arc-shaped guard plate 308 and a second arc-shaped guard plate 309 are arranged between the two rotating flower rods 302 . The first arc-shaped guard plate 308 and the second arc-shaped guard plate 309 are both fixedly connected to the first conveyor belt 101 and the second conveyor belt 102 .

[0141] A plurality of first rotating rollers 306 and a plurality of second rotating rollers 307 are arranged inside the flip box 303. The first rotating rollers 306 are rotationally connected to the flip box 303 via two first one-way bearings, and the second rotating rollers 307 are rotationally connected to the flip box 303 via two second bearings. The rotation directions of the first rotating rollers 306 and the second rotating rollers 307 are opposite, and the first rotating rollers 306 and the second rotating rollers 307 are in conflict with the corresponding pushing blocks 304 and connecting blocks 305.

[0142] The surfaces of the first rotating roller 306 and the second rotating roller 307 are not smooth and have elasticity, the inner walls of the first arc-shaped guard plate 308 and the second arc-shaped guard plate 309 are smooth and have elasticity, and the pushing block 304 and the pushing plate 310 are elastic.

[0143] In the embodiment of the present invention, the pushing block 304 on the surface of the control column 301 is operated to make the rotating flower rod 302 drive the flip box 303 to rotate counterclockwise until the extension direction of two flip boxes 303 is parallel to the conveying direction of the first conveyor belt 101, so that the PCB board on the first conveyor belt 101 enters the flip box 303, and it is also convenient for the flip box 303 close to the second conveyor belt 102 to push out the PCB board (if any). In the process of pushing the PCB board into the flip box 303, the PCB board moves inside the flip box 303. When the first rotating roller 306 is rotated, the first rotating roller 306 is driven to roll, so as to reduce the moving resistance of the PCB board, thereby reducing the wear of the PCB board. A sensor can be installed on the surface of the flip box 303 at one end where the PCB enters and exits. When the sensor senses that the PCB board has completely entered or left the inside of the flip box 303, the push block 304 is controlled to operate, so that the rotating flower rod 302 rotates a certain angle, thereby rotating the next empty flip box 303 to a position parallel to the conveying direction of the first conveyor belt 101, so that the next PCB can enter the inside of the flip box 303;

[0144] In the process of the flip box 303 following the rotating flower rod 302, since the surface of the first rotating roller 306 is not smooth and the first rotating roller 306 cannot rotate clockwise under the action of the first one-way bearing, the static friction force generated by the first rotating roller 306 on the PCB board prevents the PCB board from leaving the inside of the flip box 303 in the process of following the rotation of the flip box 303. When the flip box 303 rotates 180 degrees from the left side of the first conveyor belt 101 to the side close to the second conveyor belt 102, the PCB board is now separated from the inside of the flip box 303 by gravity. The first arc-shaped guard plate 308 is used to prevent the turning box 303 from rotating too fast, which causes the PCB board to be thrown out of the turning box 303. The elasticity of the pushing block 304 and the connecting block 305 makes it difficult for the PCB board to be rigidly damaged when it collides with the pushing block 304. The smoothness and elasticity of the first arc-shaped guard plate 308 make it difficult for the PCB board to be rigidly damaged when it collides with or slides with the first arc-shaped guard plate 308.

[0145] When the PCB board is transferred from the flip box 303 to the second conveyor belt 102, it indicates that the PCB board inside the flip box 303 close to the first conveyor belt 101 has been loaded. When the rotating flower rod 302 is controlled by the push block 304 to drive the flip box 303 to rotate at a certain angle, the empty flip box 303 rotates, and the push block 304 inside it will drive the connecting block 305 and another push block 304 on the surface of the connecting block 305 to move downward under the action of gravity. The function of the second arc guard plate 309 is to prevent the push block 304 from leaving the inside of the flip box 303 and avoid the push block 304 from interfering with the first conveyor belt 101. At the same time, under the constraint of the second bearing, the smoothness and elasticity of the second arc guard plate 309 are to reduce the wear of the push block 304 so as to extend its service life.

[0146] It should be noted that the shape of the rotating flower rod 302 consists of a cylindrical block portion and a number of rectangular rod portions fixedly connected to the surface of the cylindrical portion in a circular array, so that the corresponding two rectangular rod portions are fixedly connected to one of the flip boxes 303. The shape of the flip box 303 is rectangular, and a rectangular through groove begins to be provided on its surface to facilitate the entry and exit of the PCB board.

[0147] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A fully automatic testing equipment for PCB boards, characterized in that: It includes a detection unit, a transmission unit and a control unit; The detection unit comprises: A first image collector (201) is used to collect image information of the front side of the PCB board; A second image collector (202) is used to collect image information of the back side of the PCB board; An online X-ray scanner (203) is used to collect internal image information of the PCB board; A labeling machine (204), used for tracking and recording PCB boards; The transmission unit comprises: A first conveyor belt (101) is used to convey a PCB board so that the first image collector (201) completes an acquisition action; The second conveyor belt (102) is used to convey the PCB board so that the second image collector (202) and the online X-ray scanner (203) complete the acquisition action; the PCB flipping piece (103) is used to flip the PCB board so that the flipping action is completed when the PCB board is transferred from the first conveyor belt (101) to the second conveyor belt (102); The PCB turning member (103) comprises a plurality of turning boxes (303) arranged in a ring array between the first conveyor belt (101) and the second conveyor belt (102); a push block (304) is slidably connected to the inside of each of the turning boxes (303); a plurality of push plates (310) are equidistantly and fixedly connected to the surface of the first conveyor belt (101); the PCB board on the surface of the first conveyor belt (101) is pushed into the inside of the push block (304) by the push plate (310); and the PCB board is conveyed from the surface of the first conveyor belt (101) to the surface of the second conveyor belt (102) after being turned over by the push block (304) through the linkage between the two corresponding push blocks (304); The control unit comprises: A data acquisition module, used to process image information and collect and transmit the operation data of the machine tool; The data processing module eliminates the placement deviation through image alignment according to the image information, and finally outputs the image data, which is used to input the image data into the trained first learning model to identify the defect data of the PCB board and output the defect level of the PCB board; An action adjustment module is used to screen the PCB for waste products at each image acquisition stage of the detection unit according to the defect level, obtain a waste skip rate, so as to reduce excessive use of the PCB flipping piece (103), and dynamically adjust the speed of the first conveyor belt (101) and the second conveyor belt (102) according to the real-time change of the waste skip rate, so as to control the comprehensive cost during the equipment detection process; The defect location module is used to generate a defect location map based on the defect data of the PCB board; The first learning model after training is obtained specifically through the following steps: Collect several groups of historical test data as sample sets, and perform data enhancement on the sample sets to obtain enhanced sample sets; The enhanced sample set is randomly divided into a training set, a validation set, and a test set, the image data in the training set is used as the input of the first learning model, the defect level in the training set is used as the output of the first learning model, the first learning model is trained, the training process of the first learning model is monitored through the validation set, the performance of the first learning model is evaluated through the test set, and finally the first learning model that meets the preset accuracy is output; The PCB is screened for rejects at each image acquisition stage of the inspection unit according to the defect level, and a reject skip rate is obtained to reduce excessive use of the PCB flipping part (103), including: Score the PCB at each acquisition stage, set a dynamic screening threshold, and determine the acquisition action for the next stage based on the dynamic screening threshold; Filter actions include: In the process of collecting graphic data, the defect data in the image data is identified through the trained first learning model, and the corresponding defect level is output. If the PCB is marked as scrap, the subsequent stages are skipped; The speeds of the first conveyor belt (101) and the second conveyor belt (102) are dynamically adjusted according to the real-time change of the scrap skip rate to control the comprehensive cost in the equipment detection process. The calculation formula of the comprehensive cost is: ； in, and are the speed power consumption coefficients of the first conveyor belt (101) and the second conveyor belt (102), respectively, and are the speeds of the first conveyor belt (101) and the second conveyor belt (102), respectively, and are the basic power consumptions of the first conveyor belt (101) and the second conveyor belt (102), respectively. and The friction loss caused by the operation of the first conveyor belt (101) and the second conveyor belt (102); and is the weight coefficient; is the length of a single PCB board, is the number of PCB boards to be tested in the equipment, is the minimum speed of the two conveyor belts; Generate defect location map based on PCB defect data, including: After the detection, an image is generated to mark the defect position, the generated image is stored as a file and associated with the PCB number, and the barcode attached to the corresponding PCB board by the labeling machine (204) is scanned by a scanning device, so that the terminal device displays the corresponding defect location map; The PCB turning piece (103) further comprises: A column (301) is arranged between the first conveyor belt (101) and the second conveyor belt (102), a rotating flower rod (302) is symmetrically arranged between the two columns (301), the pushing blocks (304) are fixedly connected to the two rotating flower rods (302), a pushing block (304) is installed on the surface of one of the columns (301), and an output shaft of the pushing block (304) is fixedly connected to the corresponding rotating flower rod (302); A plurality of connecting blocks (305) are symmetrically arranged between two corresponding pushing blocks (304), the connecting blocks (305) are fixedly connected to the two corresponding pushing blocks (304), and the length of the connecting blocks (305) is the sum of the length of the flip box (303) along the extension direction of the first conveyor belt (101) and the distance between the two corresponding flip boxes (303); A first arc-shaped guard plate (308) and a second arc-shaped guard plate (309) are provided between the two rotating flower rods (302), and the first arc-shaped guard plate (308) and the second arc-shaped guard plate (309) are both fixedly connected to the first conveyor belt (101) and the second conveyor belt (102); A plurality of first rotating rollers (306) and a plurality of second rotating rollers (307) are arranged inside the flip box (303); the first rotating rollers (306) are rotationally connected to the flip box (303) via two first one-way bearings; the second rotating rollers (307) are rotationally connected to the flip box (303) via two second bearings; the first rotating rollers (306) and the second rotating rollers (307) rotate in opposite directions; the first rotating rollers (306) and the second rotating rollers (307) are in conflict with the corresponding pushing blocks (304) and the connecting blocks (305); The surfaces of the first rotating roller (306) and the second rotating roller (307) are not smooth and are elastic, the inner walls of the first arc-shaped guard plate (308) and the second arc-shaped guard plate (309) are smooth and are elastic, and the pushing block (304) and the pushing plate (310) are elastic.

Images