A circuit board detection method and system based on industrial CT

Abstract

The present invention relates to the technical field of circuit board detection, and specifically to a circuit board detection method and system based on industrial CT. The technical solution of the present invention performs industrial CT scanning on a circuit board to be tested to obtain a three-dimensional model of the circuit board to be tested; the three-dimensional model is decomposed into a plurality of units to be tested according to the information recorded in the BOM table of the circuit board to be tested, and the structure represented by the unit to be tested includes at least a solder joint structure, and each unit to be tested is scaled to a target ratio suitable for user visual inspection, and the structure represented by each unit to be tested is subjected to external morphological detection and / or internal sectioning detection at the target ratio to obtain the component detection result of each unit to be tested; all component detection results are input into the detection column of the corresponding component in the BOM table to obtain the CT detection result of the circuit board to be tested. This technical solution effectively solves the problem that the number of solder joints on the circuit board is dense and difficult to detect, and improves the accuracy of the three-dimensional model detection of the circuit board CT scanning.

Description

Technical Field

[0001] The present invention relates to the technical field of circuit board detection, and in particular to a circuit board detection method and system based on industrial CT. Background Art

[0002] For circuit board inspection, it is advantageous to scan a three-dimensional model (or stereo model) using industrial CT (Computed Tomography) for inspection. Compared with two-dimensional images, three-dimensional models can provide more comprehensive inspection results. Taking solder joints as an example, more information can be obtained through the three-dimensional model of solder joints, including the height and shape of the solder joints, which can help users observe the solder joints from different angles. Some defects that are difficult to detect from a two-dimensional perspective, such as voids inside the solder joints and insufficient welding, can be more easily detected in the three-dimensional model.

[0003] However, the dense arrangement of solder joints on the circuit board will increase the difficulty of visual inspection for users. On the one hand, due to the large number and density of solder joints, a large amount of detailed information will be presented in the 3D model, which can easily cause visual confusion and make it difficult for inspectors to quickly and accurately locate solder joints that may have defects. On the other hand, the mutual occlusion between dense solder joints will also affect the complete observation of solder joints. Some solder joint defects hidden inside or partially obscured by other solder joints are difficult to be directly discovered. Summary of the invention

[0004] In order to solve the technical problem of how to accurately detect the three-dimensional model of the circuit board CT scanning, the purpose of the present invention is to provide a circuit board detection method and system based on industrial CT, which improves the accuracy of the circuit board CT scanning three-dimensional model detection. The technical solution adopted is as follows:

[0005] In a first aspect, an embodiment of the present invention provides a circuit board detection method based on industrial CT, the method comprising:

[0006] Perform industrial CT scanning on the circuit board to be tested to obtain a three-dimensional model of the circuit board to be tested;

[0007] Performing a decomposition task on the three-dimensional model according to the information recorded in the BOM table of the circuit board to be tested, so as to decompose the three-dimensional model into a plurality of units to be tested, wherein each unit to be tested includes at least one component of the circuit board to be tested;

[0008] Each unit to be inspected is scaled to a target ratio suitable for visual inspection by a user, and an external morphology inspection and / or an internal section inspection is performed on the structure represented by each unit to be inspected at the target ratio to obtain a component inspection result of each unit to be inspected, wherein the structure represented by the unit to be inspected includes at least a solder joint structure;

[0009] Enter the test results of all components into the test column of the corresponding components in the BOM table to obtain the CT test results of the circuit board to be tested.

[0010] In an optional embodiment, a decomposition task is performed on the three-dimensional model according to information recorded in the BOM table of the circuit board to be tested, so as to decompose the three-dimensional model into a plurality of units to be tested, including:

[0011] Mark the position of the component sub-model of the 3D model according to the component number recorded in the BOM table to obtain the coordinate data of each component sub-model;

[0012] According to the component specification data recorded in the BOM table, determine the independent segmentation value of each component sub-model;

[0013] The three-dimensional model is decomposed into a plurality of units to be inspected according to the coordinate data and the independent segmentation value, wherein each component sub-model is located in a corresponding unit to be inspected.

[0014] In an optional embodiment, the component specification data includes package type, external dimension data and number of pins; and according to the component specification data recorded in the BOM table, determining the independent segmentation value of each component sub-model includes:

[0015] When the pin density of the component represented by the package type is greater than the density threshold, and the number of pins is greater than the number threshold, determining the independent segmentation value of the corresponding component sub-model to be the maximum extreme value;

[0016] When the pin density of the component represented by the package type is less than or equal to the density threshold, the independent segmentation value of each component sub-model is determined according to the package type, the shape dimension data, and the number of pins.

[0017] In an optional embodiment, determining the independent segmentation value of each component sub-model according to the package type, the external dimension data, and the number of pins includes:

[0018] Determine the top view area of ​​the corresponding component sub-model according to the shape dimension data;

[0019] Determine the target package value of the corresponding component in the corresponding package type according to a preset type value table, wherein the type value table is a correspondence table between different package types and package values;

[0020] Determine the corresponding first weight coefficient, second weight coefficient and third weight coefficient in a preset weight relationship table according to the top view area of ​​the shape, the target package value and the number of pins;

[0021] According to the formula , obtain the independent segmentation value W of each component sub-model t, where w1 is the top view area of ​​the shape, k1 is the first weight coefficient, w2 is the target package value, k2 is the second weight coefficient, w3 is the number of pins, and k3 is the third weight coefficient.

[0022] In an optional embodiment, decomposing the three-dimensional model into a plurality of units to be inspected according to the coordinate data and the independent segmentation values ​​includes:

[0023] Input all coordinate data into the distance calculation model to obtain the relative distance between each component sub-model;

[0024] Perform clustering on the relative distances between the sub-models of each component to obtain multiple sub-model clusters;

[0025] When the independent segmentation value of the component sub-model in the sub-model cluster is greater than a preset value, the corresponding component sub-model is determined as an independent unit to be inspected;

[0026] When the independent segmentation value of the component sub-model in the sub-model cluster is less than or equal to the preset value, all the component sub-models in the corresponding sub-model cluster are determined as mixed units to be inspected.

[0027] In an optional embodiment, before determining the independent unit to be inspected or the mixed unit to be inspected, the method further includes:

[0028] When the number of component sub-models in a sub-model cluster is greater than a preset number threshold, it is determined as a target sub-model cluster;

[0029] Perform mean calculation and variance calculation on the independent segmentation values ​​of all component sub-models in the target sub-model cluster to obtain the segmentation mean and segmentation variance of the target sub-model cluster;

[0030] According to the segmentation mean and segmentation variance, the preset value of the corresponding target sub-model cluster is determined.

[0031] In an optional embodiment, scaling each unit to be inspected to a target ratio suitable for visual inspection by a user includes:

[0032] Determine the inspection items of each unit to be inspected according to the corresponding information recorded in the BOM table;

[0033] According to the user's visual inspection window specifications, the corresponding target ratio is configured for the items to be inspected of each unit to be inspected, so that the user can perform visual inspection on the items to be inspected at the target ratio.

[0034] In an optional embodiment, performing external morphology detection and / or internal section detection on the structure represented by each unit to be inspected at a target ratio to obtain a component detection result of each unit to be inspected includes:

[0035] At the target ratio, determine whether the structure to be inspected represented by each unit to be inspected can be fully displayed within the preset window specifications;

[0036] If yes, perform external morphology detection and / or internal section detection on the structure represented by the first unit to be inspected that is fully displayed under the window specification to obtain component detection results of the first unit to be inspected;

[0037] If not, multiple display windows are configured for the second unit to be inspected that is not fully displayed under the window specifications, so that each display window displays the corresponding area of ​​the second unit to be inspected for external morphology detection and / or internal section detection to obtain the component detection result of the second unit to be inspected.

[0038] In an optional embodiment, before performing the decomposition task on the three-dimensional model according to the information recorded in the BOM table of the circuit board to be tested, the method further includes:

[0039] Take a picture of the circuit board to be tested, and input the captured image into a preset convolutional neural network model;

[0040] According to the output results of the convolutional neural network model, the BOM table of the circuit board to be tested is obtained.

[0041] In a second aspect, an embodiment of the present invention further provides a circuit board detection system based on industrial CT, the system comprising:

[0042] A scanning terminal is used to perform industrial CT scanning on the circuit board to be tested to obtain a three-dimensional model of the circuit board to be tested;

[0043] A detection terminal connected to the scanning terminal; the detection terminal is used to perform a decomposition task on the three-dimensional model according to the information recorded in the BOM table of the circuit board to be tested, so as to decompose the three-dimensional model into a plurality of units to be tested, wherein each unit to be tested includes at least one component of the circuit board to be tested;

[0044] The detection terminal is also used to scale each unit to be inspected to a target ratio suitable for visual inspection by a user, and perform external morphology detection and / or internal section detection on the structure represented by each unit to be inspected at the target ratio to obtain a component detection result of each unit to be inspected, wherein the structure represented by the unit to be inspected includes at least a solder joint structure;

[0045] The detection terminal is also used to input the detection results of all components into the detection column of the corresponding components in the BOM table to obtain the CT detection result of the circuit board to be tested.

[0046] The present invention has the following beneficial effects:

[0047] The technical solution of the present invention performs industrial CT scanning on the circuit board to be tested to obtain a three-dimensional model of the circuit board to be tested; performs a decomposition task on the three-dimensional model according to the information recorded in the BOM table of the circuit board to be tested, decomposes the three-dimensional model with dense solder joints into multiple units to be tested that are easier to detect, and each unit to be tested includes at least one component of the circuit board to be tested; the structure represented by the unit to be tested includes at least a solder joint structure, and each unit to be tested is scaled to a target ratio suitable for user visual inspection, so that the densely distributed solder joints can be carefully observed under a conventional viewing angle, and the structure represented by each unit to be tested is subjected to external morphological detection and / or internal section detection at the target ratio, and the component detection result of each unit to be tested is obtained to determine whether there are defects such as cold solder joints and insufficient solder filling in the solder joints; all component detection results are input into the detection column of the corresponding components in the BOM table, so that the detection results are accurately associated with each component on the circuit board, and the CT detection result of the circuit board to be tested is comprehensively and accurately obtained. The technical solution effectively solves the problem that the number of solder joints on the circuit board is dense and difficult to detect, and enables the complex three-dimensional model of CT scanning to be accurately analyzed in a process-based manner, thereby improving the accuracy of the three-dimensional model detection of the circuit board CT scanning. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0049] Figure 1 A flowchart of a circuit board detection method based on industrial CT provided by an embodiment of the present invention;

[0050] Figure 2 A schematic diagram of a three-dimensional model of all solder joints arranged on a PCB provided by an embodiment of the present invention;

[0051] Figure 3 A schematic diagram of the distribution of welding points in a unit to be inspected provided by one embodiment of the present invention;

[0052] Figure 4 A schematic structural diagram of a circuit board detection system based on industrial CT provided by an embodiment of the present invention.

[0053] Explanation of the reference numerals: 1-welding point, 2-air hole. DETAILED DESCRIPTION

[0054] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation, structure, features and effects of a circuit board detection method and system based on industrial CT proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0055] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0056] See also Figure 1 , Figure 1 A flowchart of a circuit board detection method based on industrial CT provided by an embodiment of the present invention. The detection method can be run based on a post-processing terminal of an industrial CT scanning device. The post-processing terminal can be a computer or a server, for example, running as a plug-in on a computer that generates a three-dimensional model. The terminal can run the detection method, and no specific restrictions are made here. The detection method includes:

[0057] S11. Perform industrial CT scanning on the circuit board to be tested to obtain a three-dimensional model of the circuit board to be tested.

[0058] Specifically, the circuit board to be tested is a circuit board on which components have been soldered to a PCB (Printed Circuit Board). The same scanning parameters can be used for solder joints and components, or different scanning parameters can be used. If the same scanning parameters are used, it is necessary to select a parameter setting that can take into account the imaging quality of solder joints and components. For example, in terms of resolution, to ensure that the shape and size of solder joints and details such as the outline and pins of components can be clearly distinguished, a higher spatial resolution is required. The applicable resolution can be determined based on the minimum size of solder joints and components. For scenarios where the minimum solder joint diameter is 0.3mm and the minimum component pin width is 0.2mm, the scanning resolution is set to 0.1mm.

[0059] If different scanning parameters are used, for solder joints, due to their relatively small size and the focus on the integrity of the connection, more emphasis needs to be placed on high resolution and appropriate contrast to highlight the shape and internal structure of the solder joints (such as whether there are pores, etc.). The pixel sampling rate of the scan can be appropriately increased to improve the resolution, while adjusting the radiation energy and detector sensitivity to enhance the contrast between the solder joints and the surrounding materials. For components, attention needs to be paid to their outer contours and internal structures (such as the internal packaging of the chip, etc.), and the scanning parameters can be adjusted according to the size of the components and the complexity of the internal structure. For example, for large electrolytic capacitors, a lower scanning speed and higher radiation energy may be required to ensure that the outer shell can be penetrated and the information of the internal electrodes can be obtained.

[0060] Exemplarily, step S11 includes sub-steps S11-1 to S11-3, which are specifically described as follows:

[0061] S11-1. Configure the scanning parameters of the industrial CT scanning equipment, including the scanning angle and the number of scanning layers. In order to ensure the integrity of data acquisition during the CT scanning process and to accurately construct a three-dimensional model of the circuit board to be tested, it is necessary to ensure that the rays can cover the entire periphery of the circuit board to be tested during the scanning process to avoid scanning blind spots. This can be achieved by adjusting the scanning angle and the number of scanning layers. For example, a 360° full-angle scan is used, and a sufficient number of scanning layers is set according to the overall thickness of the circuit board to be tested to ensure that information from the surface to all levels of the circuit board to be tested can be obtained.

[0062] S11-2. Perform a CT scan on the circuit board to be tested to obtain scanning data. During the scanning process, the industrial CT scanning equipment needs to be calibrated to ensure that the relative positions of the radiation source, detector and circuit board to be tested are accurate. At the same time, for multiple scans (such as multi-angle scanning), data alignment is required to accurately splice the data obtained from scanning at different angles and positions. This can be achieved by setting marking points on the circuit board to be tested or using a high-precision positioning device. For example, metal marking points are set at the four corners of the circuit board. The metal marking points have obvious characteristics in the scan data, and the scan data is aligned by identifying these metal marking points.

[0063] S11-3, use an applicable image reconstruction algorithm to construct a three-dimensional model. The tomographic image of the circuit board to be tested can be reconstructed based on the collected scanning data based on the filtered back projection algorithm, and on this basis, a three-dimensional model can be constructed through the lamination technology. At the same time, the iterative reconstruction algorithm can better handle noise and artifacts and improve the construction quality of the three-dimensional model. By comparing the model effects reconstructed by different algorithms, a reconstruction algorithm suitable for the circuit board to be tested is selected, and the parameters of the algorithm, such as the number of iterations, filtering parameters, etc., are adjusted according to the actual situation to ensure that the components and solder joints on the circuit board to be tested can be carefully portrayed. After obtaining the three-dimensional model of the circuit board to be tested, enter step S12.

[0064] S12, performing a decomposition task on the three-dimensional model according to the information recorded in the BOM table of the circuit board to be tested, so as to decompose the three-dimensional model into a plurality of units to be tested, wherein each unit to be tested includes at least one component of the circuit board to be tested.

[0065] Specifically, the BOM (or bill of materials) contains detailed information about all components on the circuit board to be tested, such as the model, specification, location, quantity, etc. of the components. The BOM can be used to determine the number of solder joints of each component. The BOM can be obtained based on the design data of the circuit board to be tested, or by reading the marking characters on the components and filling in a table.

[0066] In practical applications, since the BOM tables of different circuit boards to be tested may have different formats, there is a problem of insufficient applicability in performing the decomposition task of the three-dimensional model based on the BOM table, and there may be information errors in the manually established BOM table. In a specific implementation, the BOM table is constructed based on a convolutional neural network model, and the method specifically includes:

[0067] Take a picture of the circuit board to be tested, and input the image into the preset convolutional neural network model. The way to take a picture of the circuit board to be tested can be selected based on actual needs. For example, when there are components soldered on the front of the circuit board to be tested, the surface of the circuit board to be tested with components soldered can be photographed from multiple angles, including oblique side shots and overhead shots. The photographed image is input into the convolutional neural network model to implement component recognition, and the output result of the convolutional neural network model is filled into the template of the BOM table to obtain the BOM table of the circuit board to be tested.

[0068] It should be noted that when implementing component recognition based on the convolutional neural network model, the convolutional neural network model can be trained first until its prediction accuracy reaches the preset target; then the image taken of the circuit board to be tested is input into the convolutional neural network model for component recognition. When implementing recognition based on the convolutional neural network model, it is sufficient to be able to identify the package type, material name and number of pins of the components. The recognition method of the convolutional neural network model can effectively improve the efficiency and accuracy of BOM construction, make the obtained BOM format unified, and can be applied to the three-dimensional model to perform decomposition tasks. Please refer to Table 1, which is a schematic diagram of some components of the BOM.

[0069] Table 1:

[0070]

[0071] As can be seen from Table 1, the BOM table can be used to determine the package type, specification, number of pins and other information of each component, and the three-dimensional model can be divided based on the information recorded in the BOM table. When extracting information from the BOM table, the data of the three-dimensional model can be read. The data of the three-dimensional model contains the location information and density information of the pixel points (or voxels). The data of the BOM table and the three-dimensional model can be parsed through user operations or software tools to establish a corresponding relationship between the components in the BOM table and the sub-models of the three-dimensional model. For example, the component number in the BOM table can be identified through auxiliary software, and the corresponding component sub-model can be found in the three-dimensional model through position or shape characteristics, so that a corresponding relationship is established between the two.

[0072] After completing the establishment of the corresponding relationship between components and component sub-models in the BOM table, the units to be inspected can be divided according to the physical connection and functional relationship of the components. For example, for a filter circuit composed of multiple capacitors and resistors, these components work together in circuit functions and can be divided into the same unit to be inspected. The distribution of solder joints can also be considered. If a group of components are connected together through common solder joints, these components and solder joints can constitute a unit to be inspected. For example, the connection part between the pins of an integrated circuit chip and the PCB, including the chip pins, solder joints, and the printed circuits connected to them, can be divided into a unit to be inspected.

[0073] See also Figure 2 , Figure 2The figure is a schematic diagram of the three-dimensional model of all solder joints arranged on the PCB. It can be seen from the figure that the distribution of solder joints in various areas on the PCB is not uniform, and there are differences in the size of solder joints. When the solder joints are not decomposed, the three-dimensional model needs to be flipped and translated multiple times when implementing solder joint detection, and detection confusion and missed detection are prone to occur during the operation. Therefore, it is necessary to decompose the three-dimensional model and decompose the three-dimensional model into multiple units to be inspected according to the set division rules. For example, each unit to be inspected can be divided by drawing a virtual boundary in the three-dimensional model. The drawn boundary can be determined based on factors such as the outline of the components and the distribution of the solder joints, and no specific restrictions are made here.

[0074] It should be noted that during the decomposition process, the data integrity of each unit to be inspected must be ensured. That is, the three-dimensional data structure of each unit to be inspected (including the structure contained in the external shape and internal features) can be completely extracted for subsequent independent inspection. For example, for a unit to be inspected containing a multi-layer circuit board structure, it is necessary to ensure that the information of all layers is included in the unit to be inspected. The three-dimensional model can be decomposed into a solder joint sub-model and a component sub-model, and further decomposition is implemented based on the distribution characteristics of the solder joint sub-model; for the components that need to be inspected, the component sub-model is decomposed.

[0075] It can be understood that when decomposing the three-dimensional model, the three-dimensional model can be decomposed based on a decomposition tool (such as a split body tool), and each decomposed unit to be inspected can be stored accordingly. After the decomposition of the three-dimensional model is completed, each decomposed unit to be inspected is marked, and the marked information may include the unit number, the model of the main components contained, etc. For example, it is marked as "unit to be inspected 1-contains filter capacitors C1, C2 and resistor R1". Record the relevant attribute information of each unit to be inspected, such as the size (size range) and position (relative position coordinates in the circuit board) of the unit to be inspected. These records will help to quickly locate and analyze the test results of each unit to be inspected in the subsequent detection process.

[0076] Exemplarily, step S12 includes sub-steps S12-1 to S12-3, which are specifically described as follows:

[0077] S12-1. Mark the position of the component sub-model of the three-dimensional model according to the component number recorded in the BOM table to obtain the coordinate data of each component sub-model. The component number represents the identity information of the component on the circuit board to be tested. Each component has a unique component number and each component has corresponding external dimensions. A coordinate system can be constructed based on the top view of the three-dimensional model. The origin of the coordinate system can be located at the corner of the PCB board. The component number is marked at the geometric center of the corresponding component. The coordinate data of each component sub-model is obtained based on this. The relative spacing between components can be obtained through the coordinate data.

[0078] S12-2. Determine the independent segmentation value of each component sub-model based on the component specification data recorded in the BOM table. The component specification data includes package type, number of pins, external dimension data, etc., and the independent segmentation value is calculated based on this. The independent segmentation value represents the weight of the independent segmentation of the component sub-model. When the independent segmentation value is greater than the segmentation threshold, it means that the independent segmentation value is large, and the component sub-model needs to be segmented into independent detection units; conversely, when the independent segmentation value is less than or equal to the segmentation threshold, the component sub-model can be divided into the same unit to be inspected with other component sub-models.

[0079] In practical applications, since the number and density of pins of components directly affect the difficulty of detection, the calculation of independent segmentation values ​​can be implemented accordingly. Sub-step S12-2 includes S12-2-1 and S12-2-2, which are specifically described as follows:

[0080] S12-2-1. When the pin density of the component represented by the package type is greater than the density threshold, and the number of pins is greater than the number threshold, it means that the number of pins of the component sub-model is large and densely distributed, and it needs to be divided into independent units to be tested for testing, then the independent segmentation value of the corresponding component sub-model is determined to be the maximum extreme value. For example, when the package type is BGA (Ball Grid Array) or QFP (Quad Flat Package), it means that the pin density of the component is greater than the density threshold. If the number of pins is greater than the number threshold at the same time, the independent segmentation value of the component sub-model is assigned the maximum extreme value so that it can be independently divided into units to be tested later.

[0081] S12-2-2. When the pin density of the component represented by the package type is less than or equal to the density threshold, the independent segmentation value of each component sub-model is determined based on the package type, the external dimension data, and the number of pins. Please continue to refer to Table 1. The external dimension data of the component can be parsed through the specification / model information of the BOM table, and normalized based on the package type, external dimension data and the number of pins, so that the normalized results represented by the package type, external dimension data and the number of pins are all between 0 and 1, and the independent segmentation value is calculated based on the cumulative sum of the normalized results.

[0082] Since the package type, the external dimensions, and the number of pins characterize the detection difficulty of components from different dimensions, in order to further improve the accuracy of the independent segmentation value calculation, the embodiment of the present invention proposes the following implementation method:

[0083] The first step is to determine the top-view area of ​​the corresponding component sub-model based on the external dimension data. The external dimension data is based on the specifications of the component. For example, when the chip is a QFP package with 100 pins, based on its data sheet, it can be obtained that the package size is 14mm×14mm, the pin spacing is 0.5mm, and the top-view area is 196mm. 2 The top-down area represents the area occupied by the corresponding component sub-model on the PCB from a top-down perspective.

[0084] The second step is to determine the target package value of the corresponding component in the corresponding package type according to the preset type value table, where the type value table is a correspondence table between different package types and package values. The type value table can be set based on actual needs, and its purpose is to digitize the package type. For example, the package value of the QFP package is defined as 0.7, and the package value of the BGA package is defined as 0.9. Based on the package type of each component in the BOM table, the corresponding target package value is determined in the type value table.

[0085] The third step is to determine the corresponding first weight coefficient, second weight coefficient and third weight coefficient in the preset weight relationship table according to the top view area, target package value and number of pins. The weight relationship table is used to characterize the relationship between different component specification data and corresponding weight coefficients. The construction of the weight relationship table can be implemented based on more samples, and the sum of the first weight coefficient, the second weight coefficient and the third weight coefficient can be set to 1.

[0086] The fourth step is to use the formula , obtain the independent segmentation value W of each component sub-model t , where w1 is the top view area, k1 is the first weight coefficient, w2 is the target package value, k2 is the second weight coefficient, w3 is the number of pins, and k3 is the third weight coefficient. .

[0087] So far, the coordinate data and independent segmentation values ​​of each component sub-model have been obtained.

[0088] S12-3. Decompose the three-dimensional model into a plurality of units to be inspected according to the coordinate data and the independent segmentation values, wherein each component sub-model is located in a corresponding unit to be inspected.

[0089] Specifically, the distance between each component sub-model in the three-dimensional model can be determined through the coordinate data. If the distance is short, it is appropriate to divide it into the same unit to be inspected for inspection; on the contrary, if the distance is far, if it is divided into the same unit to be inspected for inspection, it will need to be flipped and translated for inspection due to its large volume, so it is not appropriate to divide it into the same unit to be inspected. Similarly, when the independent division value is large, it is necessary to divide the component sub-model independently into the unit to be inspected; when the independent division value is small, it is determined in combination with the coordinate data whether to divide multiple component sub-models into the same unit to be inspected, so that each unit to be inspected is easy to detect.

[0090] Further, sub-step S12-3 includes S12-3-1 to S12-3-4, which are specifically described as follows:

[0091] S12-3-1. Input all coordinate data into the distance calculation model to obtain the relative distance between each component sub-model. The relative distance can be the geometric center distance between the component sub-models; of course, it can also be the space interval distance between the component sub-models. Any distance that can characterize the relative distance between each component sub-model is sufficient, and no specific limitation is made here.

[0092] S12-3-2. Cluster the relative distances between the component sub-models to obtain multiple sub-model clusters, each sub-model cluster including at least multiple component sub-models. The distribution density presented by each component sub-model is not the same. Based on the clustering algorithm, the relative distances can be clustered, and the component sub-models with similar distances can be divided into the same sub-model cluster to obtain multiple sub-model clusters. It should be noted that when implementing the clustering process, the corresponding parameters of the clustering model can be configured based on actual needs so that the multiple sub-model clusters obtained meet the requirements of the clustering process.

[0093] S12-3-3. When the independent segmentation value of the component sub-model in the sub-model cluster is greater than the preset value, it means that the component sub-model needs to be independently segmented, and the corresponding component sub-model is determined as an independent unit to be inspected; otherwise, it is determined as a mixed unit to be inspected. The preset value can be set based on actual needs, and it can accurately segment the component sub-model that needs to be inspected separately.

[0094] In practical applications, if all sub-model clusters are set to the same preset value, it may cause the problem of insufficient applicability. Based on this, in a specific implementation, before determining the independent unit to be inspected or the mixed unit to be inspected, the method further includes:

[0095] When the number of component sub-models in a sub-model cluster is greater than a preset number threshold, it indicates that the number of component sub-models in this type of sub-model cluster is large, and it is determined as a target sub-model cluster. The mean and variance calculations are performed on the independent segmentation values ​​of all component sub-models in the target sub-model cluster to obtain the segmentation mean and segmentation variance of the target sub-model cluster; the mean and variance calculations of the independent segmentation values ​​can be calculated based on the mean value formula and the variance formula, according to the formula: , calculate the preset value J of the target sub-model cluster, where μ is the segmentation mean, s is the segmentation variance, and n is the preset coefficient. By determining the preset value in the above manner, the determined preset value can be applied to the segmentation of the corresponding target sub-model cluster, making the sub-model division of the components in the unit to be inspected more accurate.

[0096] S12-3-4. When the independent segmentation value of the component sub-model in the sub-model cluster is less than or equal to the preset value, it means that the independent segmentation values ​​of the component sub-model in the sub-model cluster are relatively similar. Each component sub-model can be allocated in the sub-model cluster, and all component sub-models in the corresponding sub-model cluster are determined as mixed units to be inspected.

[0097] At this point, the decomposition task of the three-dimensional model has been completed, and the three-dimensional model has been decomposed into multiple units to be inspected.

[0098] S13, scaling each unit to be inspected to a target ratio suitable for user visual inspection, and performing external morphology inspection and / or internal section inspection on the structure represented by each unit to be inspected at the target ratio to obtain component inspection results for each unit to be inspected, wherein the structure represented by the unit to be inspected includes at least a solder joint structure.

[0099] Specifically, the scaling ratio applicable to the components of each unit to be inspected is different, so it is necessary to implement scaling for each unit to be inspected separately. The scaling ratio applicable to each unit to be inspected under visual inspection is its target ratio, which is convenient for users to implement visual inspection. Figure 2 Taking solder joint structure inspection as an example, the solder joint structure of components packaged with DIP structure is relatively large, so visual inspection can be carried out by appropriately magnifying it; while the solder joint structure of components packaged with BGA is relatively small, so a larger magnification ratio is required for visual inspection. Therefore, it is necessary to implement corresponding target scale scaling for each unit to be inspected. At the target scale, the unit to be inspected can clearly characterize its structure so as to implement external morphology inspection and / or internal section inspection.

[0100] See also Figure 3 , Figure 3The figure is a schematic diagram of the distribution of solder joints in the unit to be inspected. For the inspection of solder joints, it is necessary to have a full shape and no pores inside. Therefore, it is necessary to perform external morphology inspection and internal section inspection on the solder joints. The solder joints can be sectioned based on software tools to detect their internal structure. As can be seen from the figure, after the solder joints are sectioned, if there are pores 2 in solder joint 1, the characteristics of the pores can be clearly observed through visual inspection. Based on the visual inspection of each unit to be inspected, the corresponding component inspection results can be obtained.

[0101] It should be noted that the tests required for different components are different. For example, some components only require solder joint testing, but for components with higher requirements for operational reliability, further testing of the components themselves is required. For example, solid-state capacitors can obtain more comprehensive CT test results by testing the components themselves.

[0102] Exemplarily, step S13 includes sub-steps S13-1 to S13-5, which are specifically described as follows:

[0103] S13-1, according to the corresponding information recorded in the BOM table of each unit to be inspected, determine the items to be inspected for each unit to be inspected. By reading the information recorded in the BOM table of each unit to be inspected, the components contained in each unit to be inspected can be determined, and the items to be inspected are determined based on the inspection requirements of the components, including solder joint structure inspection, component internal inspection, etc.

[0104] S13-2. Configure the corresponding target ratio for the items to be inspected of each unit to be inspected according to the window specifications for visual inspection by the user, so that the user can perform visual inspection on the items to be inspected at the target ratio. The window specifications are the window size for visual inspection by the user based on a display tool (such as a display), which can be obtained based on the window size presented by the display tool. The target ratio for scaling the items to be inspected of each unit to be inspected is configured based on the window specifications, and the unit to be inspected can clearly display the structure represented by the unit to be inspected at the target ratio. The scaling ratios configured for different items to be inspected can be set based on actual needs, as long as they are suitable for user visual inspection under the items to be inspected.

[0105] S13-3. Determine whether the structure to be inspected represented by each unit to be inspected can be fully displayed under the preset window specifications at the target ratio. The unit to be inspected may contain multiple components. Although the structure to be inspected can be clearly seen at the target ratio, the inspection result may be partially displayed due to excessive magnification. In this state, the user still needs to move the display area within the unit to be inspected, which will cause detection confusion when moving or flipping.

[0106] Whether the structure to be inspected is fully displayed can be determined by image edge detection. For example, the image displayed in the window can be captured, and the edge of the image can be detected to determine whether the model of the unit to be inspected is fully displayed. Since the background of the display tool and the model of the unit to be inspected displayed have different grayscales, edge detection algorithms such as Sobel operator and Canny operator can be used to extract edge information in the image. If the edges detected at the edge of the window all belong to the background, that is, the edges are composed of pixels with grayscale values ​​close to the background grayscale, then it can be considered that the unit to be inspected is fully displayed in the window at the current zoom ratio; if an edge corresponding to the display grayscale of the unit to be inspected is detected at the edge of the window, it means that part of the unit to be inspected exceeds the window range and is not fully displayed.

[0107] S13-4, if yes, it means that the unit to be inspected can fully display all its structures to be inspected, and the structure represented by the first unit to be inspected that is fully displayed under the window specifications is subjected to external morphology detection and / or internal section detection to obtain the component detection result of the first unit to be inspected. The first unit to be inspected is a unit to be inspected that can fully display the structure to be inspected under the target ratio and window specifications, and the corresponding component detection result can be obtained after completing the visual inspection of the first unit to be inspected.

[0108] S13-5, if not, it means that the entire structure of the unit to be inspected cannot be displayed through a single display window, then multiple display windows are configured for the second unit to be inspected that is not fully displayed under the window specifications, so that each display window displays the corresponding area of ​​the second unit to be inspected for external morphology detection and / or internal section detection, so as to obtain the component detection result of the second unit to be inspected. Multiple display windows can be numbered respectively, and each display window can be detected one by one. By implementing visual inspection in this way, there is no need to perform zooming operations during the detection process. Each display window can be detected one by one based on the detection process, and the detected results are determined as the component detection results of the second unit to be inspected.

[0109] So far, the component inspection results of each unit to be inspected have been obtained.

[0110] S14. Input the inspection results of all components into the inspection column of the corresponding components in the BOM table to obtain the CT inspection results of the circuit board to be tested.

[0111] Specifically, the BOM table includes all components of the circuit board to be tested. By inputting the test results of all components into the test column, the test results of each component based on CT can be determined, which can fully characterize the CT test results of the circuit board to be tested. It can be understood that when implementing the visual inspection of the circuit board to be tested, the traditional method requires the input of the test results based on the operation of the mouse and keyboard. The method based on the present invention can complete the input of the test results only by operating the keyboard, thereby effectively improving the inspection efficiency of the circuit board.

[0112] Based on the same technical concept as the detection method, the embodiment of the present invention also provides a circuit board detection system based on industrial CT, please refer to Figure 4 , Figure 4 40 is a schematic diagram of the structure of the detection system. The detection system includes a scanning terminal 401 and a detection terminal 402.

[0113] The scanning terminal 401 is used to perform industrial CT scanning on the circuit board to be tested, so as to obtain a three-dimensional model of the circuit board to be tested.

[0114] The detection terminal 402 is connected to the scanning terminal 401; the detection terminal 402 is used to perform a decomposition task on the three-dimensional model according to the information recorded in the BOM table of the circuit board to be tested, so as to decompose the three-dimensional model into multiple units to be tested, wherein each unit to be tested includes at least one component of the circuit board to be tested. The detection terminal 402 is also used to scale each unit to be tested to a target ratio suitable for user visual inspection, and perform external morphological detection and / or internal section detection on the structure represented by each unit to be tested at the target ratio to obtain the component detection result of each unit to be tested, wherein the structure represented by the unit to be tested includes at least the solder joint structure. The detection terminal 402 is also used to input all component detection results into the detection column of the corresponding component in the BOM table to obtain the CT detection result of the circuit board to be tested.

[0115] The technical solution provided in the embodiments of the present invention has at least the following technical effects or advantages:

[0116] The technical solution performs industrial CT scanning on the circuit board to be tested to obtain a three-dimensional model of the circuit board to be tested; performs a decomposition task on the three-dimensional model according to the information recorded in the BOM table of the circuit board to be tested, decomposing the three-dimensional model with dense solder joints into multiple units to be tested that are easier to detect, and each unit to be tested includes at least one component of the circuit board to be tested; the structure represented by the unit to be tested includes at least the solder joint structure, and each unit to be tested is scaled to a target ratio suitable for user visual inspection, so that the densely distributed solder joints can be carefully observed under a conventional viewing angle, and the structure represented by each unit to be tested is subjected to external morphological detection and / or internal section detection at the target ratio to obtain the component detection result of each unit to be tested, and determine whether the solder joint has defects such as cold solder joints and insufficient solder filling; all component detection results are input into the detection column of the corresponding components in the BOM table, so that the detection results are accurately associated with the components on the circuit board, and the CT detection results of the circuit board to be tested are comprehensively and accurately obtained. The technical solution effectively solves the problem that the number of solder joints on the circuit board is dense and difficult to detect, and enables the complex three-dimensional model of CT scanning to be accurately analyzed in a process-based manner, thereby improving the accuracy of the three-dimensional model detection of the circuit board CT scanning.

[0117] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0118] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A circuit board detection method based on industrial CT, characterized in that: The method comprises: Performing industrial CT scanning on the circuit board to be tested to obtain a three-dimensional model of the circuit board to be tested; Performing a decomposition task on the three-dimensional model according to the information recorded in the BOM table of the circuit board to be tested, so as to decompose the three-dimensional model into a plurality of units to be tested, wherein each unit to be tested includes at least one component of the circuit board to be tested; Each unit to be inspected is scaled to a target ratio suitable for visual inspection by a user, and an external morphology inspection and / or an internal section inspection is performed on the structure represented by each unit to be inspected at the target ratio to obtain a component inspection result of each unit to be inspected, wherein the structure represented by the unit to be inspected includes at least a solder joint structure; Input the detection results of all components into the detection column of the corresponding components in the BOM table to obtain the CT detection result of the circuit board to be tested; The performing a decomposition task on the three-dimensional model according to the information recorded in the BOM table of the circuit board to be tested, so as to decompose the three-dimensional model into a plurality of units to be tested, comprises: Marking the positions of the component sub-models of the three-dimensional model according to the component numbers recorded in the BOM table to obtain coordinate data of each component sub-model; Determine the independent segmentation value of each component sub-model according to the component specification data recorded in the BOM table; Decomposing the three-dimensional model into the plurality of units to be inspected according to the coordinate data and the independent segmentation value, wherein each component sub-model is located in a corresponding unit to be inspected; The component specification data includes package type, external dimension data and number of pins; the independent segmentation value of each component sub-model is determined according to the component specification data recorded in the BOM table, including: When the pin density of the component represented by the package type is greater than a density threshold, and the number of pins is greater than a number threshold, determining that the independent segmentation value of the corresponding component sub-model is a maximum extreme value; When the pin density of the component represented by the package type is less than or equal to the density threshold, determining the independent segmentation value of each component sub-model according to the package type, the shape dimension data, and the number of pins; Determining the independent segmentation value of each component sub-model according to the package type, the external dimension data, and the number of pins includes: Determine the top view area of ​​the corresponding component sub-model according to the shape dimension data; Determine the target package value of the corresponding component in the corresponding package type according to a preset type value table, wherein the type value table is a correspondence table between different package types and package values; Determine the corresponding first weight coefficient, second weight coefficient and third weight coefficient in a preset weight relationship table according to the top view area of ​​the outer shape, the target package value and the number of pins; According to the formula , obtain the independent segmentation value W of each component sub-model t , wherein w1 is the top view area of ​​the outer shape, k1 is the first weight coefficient, w2 is the target packaging value, k2 is the second weight coefficient, w3 is the number of pins, and k3 is the third weight coefficient.

2. The circuit board detection method based on industrial CT according to claim 1 is characterized in that: Decomposing the three-dimensional model into the plurality of units to be inspected according to the coordinate data and the independent segmentation values ​​comprises: Input all coordinate data into the distance calculation model to obtain the relative distance between each component sub-model; Performing clustering processing on the relative distances between the component sub-models to obtain a plurality of sub-model clusters; When the independent segmentation value of the component sub-model in the sub-model cluster is greater than a preset value, determining the corresponding component sub-model as an independent unit to be inspected; When the independent segmentation value of the component sub-model in the sub-model cluster is less than or equal to the preset value, all the component sub-models in the corresponding sub-model cluster are determined as mixed units to be inspected.

3. The circuit board detection method based on industrial CT according to claim 2 is characterized in that: Before determining the independent unit to be inspected or the mixed unit to be inspected, the method further includes: When the number of component sub-models in a sub-model cluster is greater than a preset number threshold, it is determined as a target sub-model cluster; Performing mean calculation and variance calculation on the independent segmentation values ​​of all component sub-models in the target sub-model cluster to obtain the segmentation mean and segmentation variance of the target sub-model cluster; According to the segmentation mean and the segmentation variance, a preset value corresponding to the target sub-model cluster is determined.

4. The circuit board detection method based on industrial CT according to claim 1 is characterized in that: The step of scaling each unit to be inspected to a target ratio suitable for visual inspection by a user comprises: Determine the inspection items of each unit to be inspected according to the corresponding information recorded in the BOM table of each unit to be inspected; According to the visual inspection window specifications of the user, a corresponding target ratio is configured for the items to be inspected of each unit to be inspected, so that the user can perform visual inspection on the items to be inspected at the target ratio.

5. The circuit board detection method based on industrial CT according to claim 1 is characterized in that: Performing external morphology detection and / or internal section detection on the structure represented by each unit to be inspected at the target ratio to obtain component detection results of each unit to be inspected, including: Determining whether the structure to be inspected represented by each unit to be inspected can be fully displayed within a preset window specification at the target ratio; If yes, perform external morphology detection and / or internal section detection on the structure represented by the first unit to be inspected that is fully displayed under the window specification to obtain component detection results of the first unit to be inspected; If not, multiple display windows are configured for the second unit to be inspected that is not fully displayed within the window specification, so that each display window displays the corresponding area of ​​the second unit to be inspected for external morphology detection and / or internal section detection to obtain the component detection result of the second unit to be inspected.

6. The circuit board detection method based on industrial CT according to claim 1 is characterized in that: Before performing the decomposition task on the three-dimensional model according to the information recorded in the BOM table of the circuit board to be tested, the method further includes: Taking a photo of the circuit board to be tested, and inputting the taken image into a preset convolutional neural network model; According to the output result of the convolutional neural network model, a BOM table of the circuit board to be tested is obtained.

7. A circuit board detection system based on industrial CT, characterized in that: The system is a system corresponding to any detection method of claims 1-6; the system comprises: A scanning terminal is used to perform industrial CT scanning on the circuit board to be tested to obtain a three-dimensional model of the circuit board to be tested; A detection terminal connected to the scanning terminal; the detection terminal is used to perform a decomposition task on the three-dimensional model according to the information recorded in the BOM table of the circuit board to be tested, so as to decompose the three-dimensional model into a plurality of units to be tested, wherein each unit to be tested includes at least one component of the circuit board to be tested; The detection terminal is also used to scale each unit to be inspected to a target ratio suitable for visual inspection by a user, and perform external morphology detection and / or internal section detection on the structure represented by each unit to be inspected at the target ratio to obtain a component detection result of each unit to be inspected, wherein the structure represented by the unit to be inspected includes at least a solder joint structure; The detection terminal is also used to input the detection results of all components into the detection columns of the corresponding components in the BOM table to obtain the CT detection results of the circuit board to be tested.

Images