Machine learning PCB defect automatic detection system and method

Abstract

The invention discloses a PCB defect automatic detection system and method based on machine learning, and mainly relates to the technical field of defect detection. The system comprises a deformation cloud picture sequence obtaining module used for carrying out data acquisition to obtain a first PCB bending video image frame sequence, a first flexible stress distribution diagram sequence and a first deformation cloud picture sequence; the implicit trend vector obtaining module is used for obtaining a first crack implicit trend vector; the defect feature vector storage module is used for storing the first corrected crack defect feature vector into a defect memory bank; the updated defect memory bank obtaining module is used for obtaining an updated defect memory bank; and the defect detection result obtaining module is used for obtaining a defect detection result of the to-be-detected flexible PCB. The method has the beneficial effects that the technical problem of low reliability of a detection result caused by lack of dynamic defect detection of the PCB in the prior art is solved, and the technical effect of carrying out dynamic and reliable defect detection on the PCB is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of defect detection, and in particular to a machine learning-based automatic PCB defect detection system and method. Background Art

[0002] With the development of flexible electronics technology, flexible printed circuit boards (PCBs) are widely used in wearable devices, flexible displays, and bend sensors. However, flexible PCBs are often subjected to periodic bending, folding, and torsional loads during service, which can cause microcracks or hidden structural damage, seriously affecting their electrical performance and service life. Existing flexible PCB defect detection methods are mostly static, which makes it difficult to promptly detect fine cracks caused by bending fatigue or crack evolution trends. As a result, the detection results are susceptible to local noise interference, resulting in insufficient defect identification accuracy.

[0003] The existing technology has a technical problem of lack of dynamic defect detection of PCBs, resulting in low reliability of detection results. Summary of the Invention

[0004] The present application provides a machine learning-based automatic PCB defect detection system and method, which is used to solve the technical problem in the prior art of lacking dynamic defect detection in PCBs, resulting in low reliability of detection results.

[0005] In view of the above problems, the present application provides a machine learning-based automatic PCB defect detection system and method.

[0006] In a first aspect of the present application, a machine learning-based automatic PCB defect detection system is provided, the system comprising: a deformation cloud map sequence acquisition module for applying a cyclic bending load to a flexible PCB to be tested through a bending fatigue testing machine based on preset bending load cycle parameters, and using an industrial camera, a flexible stress sensor, and a deformation sensor to collect data on the bending process when the first bending load is applied, thereby obtaining a first PCB bending video image frame sequence, a first flexible stress distribution map sequence, and a first deformation cloud map sequence; and an implicit trend vector acquisition module for performing crack time series iterative trend analysis on the first PCB bending video image frame sequence to obtain a first crack implicit trend vector. ; A defect feature vector storage module is used to perform crack defect correction on the first crack implicit trend vector according to the first flexible stress distribution map sequence and the first deformation cloud map sequence, obtain a first corrected crack defect feature vector, and store the first corrected crack defect feature vector in a defect memory bank; an updated defect memory bank acquisition module is used to perform a bending crack defect test on the flexible PCB to be tested again based on the preset bending load cycle parameters, and update the defect memory bank according to the test results to obtain an updated defect memory bank; a defect detection result acquisition module is used to perform integrated correlation verification on the updated defect memory bank to obtain a defect detection result of the flexible PCB to be tested.

[0007] In a possible implementation, the bending load cycle parameter is selected from at least one of the following parameters: bending radius, bending frequency, bending angle, number of cycles, and load amplitude.

[0008] In a possible implementation, the implicit trend vector acquisition module is used to perform the following steps: performing trend feature analysis on the first PCB bending video image frame sequence according to the crack multi-time series bandwidth set, and sorting them in ascending order according to the corresponding crack multi-time series bandwidth to obtain a crack multi-time series trend vector sequence; extracting the first crack multi-time series trend vector and the second crack multi-time series trend vector from the crack multi-time series trend vector sequence for implicit trend enhancement analysis to construct a first implicit trend enhancement matrix; using the first implicit trend enhancement matrix to perform crack time series iterative trend enhancement on the second crack implicit trend vector to obtain a first-stage crack implicit trend vector; based on the first-stage crack implicit trend vector, performing crack time series iterative trend enhancement on the third crack multi-time series trend vector in the crack multi-time series trend vector sequence, and so on to obtain the first crack implicit trend vector.

[0009] In a possible implementation, the implicit trend vector acquisition module is further used to perform the following steps: based on the specifications and models of the flexible PCB to be tested, historical crack development data mining is performed to obtain a historical crack development data set; the historical crack development data set is divided according to the crack type, and crack development time is extracted from the division results to obtain M crack development time sets, where M is a positive integer; the maximum and minimum values ​​in the M crack development time sets are respectively extracted to construct the crack multi-time series bandwidth set.

[0010] In one possible embodiment, the defect feature vector storage module is used to perform the following steps: traversing the first flexible stress distribution map sequence and the first deformation cloud map sequence to perform feature identification, and determining the first flexible stress feature vector sequence and the first deformation cloud map feature vector sequence; performing trend analysis based on the first flexible stress feature vector sequence and the first deformation cloud map feature vector sequence to determine the first flexible stress trend vector and the first deformation cloud map trend vector; and performing weighted correction on the first crack implicit trend vector based on the first flexible stress trend vector and the first deformation cloud map trend vector to obtain a first corrected crack defect feature vector.

[0011] In one possible implementation, the defect detection result acquisition module is used to perform the following steps: extracting a first corrected crack defect feature vector and a second corrected crack defect feature vector from the updated defect memory library; using the crack position as an association constraint, performing association verification using the second corrected crack defect feature vector and the first corrected crack defect feature vector; if the verification fails, removing the first corrected crack defect feature vector from the updated defect memory library to obtain a stage cleaning updated defect memory library; extracting a third corrected crack defect feature vector from the stage cleaning updated defect memory library again to perform association verification on the second corrected crack defect feature vector until reaching the last corrected crack defect feature vector of the stage cleaning updated defect memory library to obtain a target cleaning updated defect memory library; using a detection result identifier to identify the target cleaning updated defect memory library to obtain the defect detection result of the flexible PCB to be tested.

[0012] In a possible embodiment, the defect detection result acquisition module is also used to perform the following steps: extracting the first crack position center coordinates, the first crack contour area, the second crack position center coordinates and the second crack contour area of ​​the first corrected crack defect feature vector and the second corrected crack defect feature vector; calculating the center position deviation distance based on the first crack position center coordinates and the second crack position center coordinates, and judging whether the calculation result meets the preset position deviation threshold; if not, judging that the position is inconsistent; calculating the intersection-and-union ratio based on the first crack contour area and the second crack contour area, and judging whether the intersection-and-union ratio meets the preset intersection-and-union ratio threshold; if not, judging that the contour diffusion is inconsistent; when the position is inconsistent or the contour diffusion is inconsistent, the association verification result is failed verification.

[0013] A second aspect of the present application provides a machine learning-based automatic PCB defect detection method, the method comprising: applying a cyclic bending load to a flexible PCB to be tested through a bending fatigue testing machine based on preset bending load cycle parameters, and using an industrial camera, a flexible stress sensor, and a deformation sensor to collect data on the bending process when the first bending load is applied, to obtain a first PCB bending video image frame sequence, a first flexible stress distribution map sequence, and a first deformation cloud map sequence; performing crack time series iterative trend analysis on the first PCB bending video image frame sequence to obtain a first crack implicit trend vector; performing crack defect correction on the first crack implicit trend vector based on the first flexible stress distribution map sequence and the first deformation cloud map sequence to obtain a first corrected crack defect feature vector, and storing the first corrected crack defect feature vector in a defect memory library; performing a bending crack defect test on the flexible PCB to be tested again based on the preset bending load cycle parameters, and updating the defect memory library based on the test results to obtain an updated defect memory library; performing integrated correlation verification on the updated defect memory library to obtain a defect detection result of the flexible PCB to be tested.

[0014] One or more technical solutions provided in this application have at least the following technical effects or advantages: Based on preset bending load cycle parameters, this application applies a cyclic bending load to the flexible PCB under test using a bending fatigue testing machine. An industrial camera, a flexible stress sensor, and a deformation sensor are used to collect data on the bending process during the first bending load application, obtaining a first PCB bending video image frame sequence, a first flexible stress distribution map sequence, and a first deformation cloud map sequence. The first PCB bending video image frame sequence is then subjected to crack temporal iterative trend analysis to obtain a first crack implicit trend vector. The first crack implicit trend vector is then subjected to crack defect correction based on the first flexible stress distribution map sequence and the first deformation cloud map sequence to obtain a first corrected crack defect feature vector. The first corrected crack defect feature vector is stored in a defect memory bank. The flexible PCB under test is then subjected to a bending crack defect test again based on the preset bending load cycle parameters. The defect memory bank is updated based on the test results to obtain an updated defect memory bank. The updated defect memory bank is then subjected to integrated correlation verification to obtain defect detection results for the flexible PCB under test. This achieves the technical effect of improving defect detection reliability by performing multiple tests and verifications on the flexible PCB under test. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Attachment Figure 1 This is a structural diagram of a machine learning-based automatic PCB defect detection system provided by an embodiment of the present invention.

[0016] Attachment Figure 2 This is a flow chart of a method for automatic PCB defect detection based on machine learning provided by an embodiment of the present invention.

[0017] Reference numerals shown in the accompanying drawings: Deformation cloud map sequence acquisition module 11, implicit trend vector acquisition module 12, defect feature vector storage module 13, updated defect memory library acquisition module 14, defect detection result acquisition module 15. DETAILED DESCRIPTION

[0018] The present invention will be further described below in conjunction with specific examples. It should be understood that these examples are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms also fall within the scope defined by the claims attached to this application. It should be noted that the terms "including" and "having" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices.

[0019] Example 1, as shown in the attached Figure 1As shown, the present application provides a machine learning PCB defect automatic detection system, the system comprising: The deformation cloud map sequence acquisition module 11 is configured to apply a cyclic bending load to the flexible PCB under test using a bending fatigue testing machine based on preset bending load cycle parameters, and to acquire data of the bending process when the bending load is first applied using an industrial camera, a flexible stress sensor, and a deformation sensor, thereby obtaining a first PCB bending video image frame sequence, a first flexible stress distribution map sequence, and a first deformation cloud map sequence. Furthermore, the bending load cycle parameter is selected from at least one of the following parameters: bending radius, bending frequency, bending angle, number of cycles, and load amplitude.

[0020] In one possible embodiment, the preset bending load cycle parameters refer to physical load conditions pre-set by a person skilled in the art prior to conducting a bending fatigue test. These parameters are selected from at least one of the following: bending radius (curvature of the bend), bending frequency (number of bends per unit time), bending angle (bending amplitude), number of cycles (total number of repeated bends), and load amplitude (applied load magnitude). An industrial camera is used to record dynamic image information during the bending process. Flexible stress sensors and deformation sensors are embedded or attached to the surface of the flexible PCB to be tested, respectively, to measure the stress distribution and local deformation during the bending process in real time, outputting them as time-series data.

[0021] The bending fatigue testing machine applies continuous and controllable bending and resetting operations to the flexible PCB under test, simulating the repeated bending conditions of the product in actual application. When the bending load is applied for the first time, the industrial camera synchronously captures the entire bending process to generate a continuous sequence of video image frames for subsequent crack timing analysis. At the same time, the flexible stress sensor and deformation sensor measure the stress distribution state and deformation of the PCB at different positions in real time, and output them as the first flexible stress distribution map sequence and the first deformation cloud map sequence, respectively. By providing a multi-modal, time-aligned raw data foundation for subsequent modules, the comprehensiveness of the data source for subsequent crack implicit trend extraction, defect feature correction and memory library verification is achieved, thereby improving the reliability of dynamic defect detection.

[0022] An implicit trend vector obtaining module 12 is configured to perform a crack time sequence iterative trend analysis on the first PCB bending video image frame sequence to obtain a first crack implicit trend vector; Furthermore, the implicit trend vector obtaining module 12 is configured to perform the following steps: Performing trend feature analysis on the first PCB bending video image frame sequence according to the crack multi-time series bandwidth set, and sorting the frames in ascending order according to the corresponding crack multi-time series bandwidths to obtain a crack multi-time series trend vector sequence; Extracting a first crack multi-time series trend vector and a second crack multi-time series trend vector from the crack multi-time series trend vector sequence to perform implicit trend enhancement analysis and construct a first implicit trend enhancement matrix; Performing crack time series iterative trend enhancement on the second crack implicit trend vector using the first implicit trend enhancement matrix to obtain a first-stage crack implicit trend vector; Based on the first-stage crack implicit trend vector, crack time series iterative trend enhancement is performed on the third crack multi-time series trend vector in the crack multi-time series trend vector sequence, and so on to obtain the first crack implicit trend vector.

[0023] Furthermore, the implicit trend vector obtaining module 12 is further configured to perform the following steps: Based on the specifications and models of the flexible PCB to be tested, historical crack development data mining is performed to obtain a historical crack development data set; Dividing the historical crack development data set according to crack type, and extracting crack development time from the division results to obtain M crack development time sets, where M is a positive integer; The maximum value and the minimum value in the M crack development time sets are respectively extracted to construct the crack multi-time series bandwidth set.

[0024] In one possible embodiment, a multi-scale trend feature analysis is performed on the captured first PCB bending video image frame sequence according to a pre-constructed set of crack multi-series bandwidths. Specifically, the image frame sequence is analyzed at different scales based on the development and evolution duration corresponding to different types of cracks. The extracted results are sorted from small to large according to the size of the crack multi-series bandwidth to form a crack multi-series trend vector sequence. Next, an implicit trend enhancement analysis is performed on the crack multi-series trend vector sequence to determine a first implicit trend enhancement matrix. The first implicit trend enhancement matrix is ​​used to optimize the second trend vector to form a first-stage crack implicit trend vector. The module then uses the first-stage results as a reference to iteratively enhance the remaining multi-series trend vectors, gradually improving the physical consistency and temporal continuity of the trend vectors, and ultimately outputting the first crack implicit trend vector.

[0025] To ensure the accuracy of the multi-time series bandwidth set, the system mines historical crack development data for similar flexible PCBs based on the specifications of the flexible PCB being tested. This data is then divided by crack type and extracted for each type of evolution duration. The maximum and minimum values ​​are then statistically analyzed to adaptively construct a bandwidth set suitable for the current test object. By amplifying and dynamically enhancing the subtle crack variations implicit in the original video data in a time-aligned manner, the system significantly improves the sensitivity and accuracy of subsequent defect correction and identification.

[0026] In one embodiment of the present application, a first element similarity set is obtained by performing a one-to-one inner product mapping on the same type of elements in the first crack multi-time series trend vector and the second crack multi-time series trend vector using a cosine similarity calculation formula. The first element similarity set reflects the degree of similarity between the first crack multi-time series trend vector and the second crack multi-time series trend vector. Furthermore, the first element similarity set is normalized, and then the normalized value after normalization is filled into the initially empty matrix to obtain a first implicit trend enhancement matrix. The first implicit trend enhancement matrix reflects the implicit correlation between the first crack multi-time series trend vector and the second crack multi-time series trend vector. The first implicit trend enhancement matrix is ​​matrix-multiplied with the second crack implicit trend vector to suppress the noise in the crack evolution time series information, enhance the key data, and obtain the first stage crack implicit trend vector.

[0027] Preferably, based on the same obtaining principle as that of obtaining the first-stage crack implicit trend vector, the first-stage crack implicit trend vector is used to perform crack timing iterative trend enhancement on the third crack multi-time series trend vector in the crack multi-time series trend vector sequence, and similarly, the remaining crack multi-time series trend vectors in the crack multi-time series trend vector sequence are iteratively enhanced to obtain the first crack implicit trend vector.

[0028] In one embodiment of the present application, first historical crack growth data is extracted from the historical crack growth data set. Based on the crack type corresponding to the first historical crack growth data, the remaining historical crack growth data in the historical crack growth data set are matched, and the historical crack growth data with the same crack type are added to a first partition set. The first partition set is removed from the historical crack growth data set, and then, based on the same principle as for obtaining the first partition set, the remaining historical crack growth data in the historical crack growth data set are partitioned based on crack type, thereby obtaining the M partition sets. M is a positive integer.

[0029] Then, the crack development duration of each historical crack development data set in the M partitioned sets is extracted to obtain M crack development duration sets. Furthermore, the maximum and minimum values ​​in each of the M crack development duration sets are extracted to obtain an initial crack multi-time series bandwidth set. Finally, the initial crack multi-time series bandwidth sets are unioned to obtain the crack multi-time series bandwidth set.

[0030] a defect feature vector storage module 13, configured to perform crack defect correction on the first crack implicit trend vector according to the first flexible stress distribution map sequence and the first deformation cloud map sequence, obtain a first corrected crack defect feature vector, and store the first corrected crack defect feature vector in a defect memory bank; Furthermore, the defect feature vector storage module 13 is configured to perform the following steps: Traversing the first flexible stress distribution map sequence and the first deformation cloud map sequence to perform feature recognition, and determining a first flexible stress feature vector sequence and a first deformation cloud map feature vector sequence; performing trend analysis based on the first flexible stress characteristic vector sequence and the first deformation cloud map characteristic vector sequence to determine a first flexible stress trend vector and a first deformation cloud map trend vector; The first crack implicit trend vector is weightedly corrected based on the first flexible stress trend vector and the first deformation cloud map trend vector to obtain a first corrected crack defect feature vector.

[0031] In one possible embodiment, the first flexible stress feature vector sequence represents an ordered sequence of vectors consisting of local or global stress features (such as maximum stress, stress gradient, and stress concentration area) extracted from each frame of the flexible stress distribution map. The first deformation cloud map feature vector sequence represents an ordered sequence of vectors consisting of deformation amount, displacement direction, and deformation concentration extracted from each frame of the deformation cloud map. The defect memory is a database or vector library that stores multiple corrected crack defect feature vectors for subsequent similarity comparison and dynamic learning.

[0032] In one possible embodiment, the first flexible stress distribution map sequence and the first deformation cloud map sequence are first processed, identifying stress concentrations, deformation peaks, and their spatial distribution frame by frame. Physical features are extracted from each frame, and these are aggregated to form a first flexible stress feature vector sequence and a first deformation cloud map feature vector sequence. Next, time series trend analysis is performed on each of these feature vector sequences to obtain overall trend vectors reflecting the stress and deformation evolution throughout the bending cycle, referred to as the first flexible stress trend vector and the first deformation cloud map trend vector. These two trend vectors are then combined with the previously obtained first crack implicit trend vector. A frame-level correction is performed on the original trend vector using a preset weighting coefficient (which can be a fixed value or adaptively learned from historical data), resulting in the output of a first corrected crack defect feature vector. Finally, this corrected feature vector, along with metadata (such as model, load parameters, and timestamp), is stored in a defect memory database, providing traceable data support for subsequent defect evolution analysis, automatic similarity retrieval, and multi-round verification.

[0033] In one possible embodiment, multiple sample flexible stress distribution maps and multiple sample flexible stress feature vectors are obtained as training data. The training data is used to perform supervised training on a framework constructed based on a feedforward neural network. During the training process, the network parameters of the framework are revised based on the accuracy of the framework output data to obtain the revised feature identifier. The feature identifier is used to perform feature recognition on the first flexible stress distribution map sequence to obtain the first flexible stress feature vector sequence. Based on the same acquisition principle as that for obtaining the first flexible stress feature vector sequence, feature recognition is performed on the first deformation cloud map sequence to obtain the first deformation cloud map feature vector sequence.

[0034] Next, a trend analyzer is obtained. The trend analyzer is a functional module constructed using a convolutional neural network framework and employing multiple sample feature vector sequences and multiple sample trend vectors as training data sets. The trend analyzer is used to perform trend analysis on the first flexible stress feature vector sequence and the first deformation cloud feature vector sequence, respectively, to obtain a first flexible stress trend vector and a first deformation cloud trend vector. Furthermore, the first flexible stress trend vector and the first deformation cloud trend vector are used to perform a weighted correction on the first crack implicit trend vector according to weights predefined by those skilled in the art, to obtain a first corrected crack defect feature vector.

[0035] By coupling and correcting the purely visual crack trend results with actual physical field data, the technical effect of greatly improving the physical consistency of crack detection and the reliability of subsequent identification is achieved.

[0036] An updated defect memory library obtaining module 14 is configured to perform a bending crack defect test on the flexible PCB to be tested again based on the preset bending load cycle parameters, and update the defect memory library according to the test results to obtain an updated defect memory library; In one possible embodiment, based on the same principle as obtaining the first corrected crack defect feature vector, a bending crack defect test is performed on the flexible PCB under test based on the preset bending load cycle parameters to obtain a corresponding corrected crack defect feature vector until the number of cycles meeting the preset bending load cycle parameters is reached. The test results, i.e., the corrected crack defect feature vector obtained from each test, are stored in a defect memory bank and updated to obtain the updated defect memory bank. By continuously accumulating multiple rounds of crack evolution data for this PCB model, a high-quality, dynamically evolving defect memory bank is formed, providing data support for subsequent integrated correlation verification and final defect identification.

[0037] The defect detection result obtaining module 15 is used to perform integrated correlation verification on the updated defect memory library to obtain defect detection results of the flexible PCB to be tested.

[0038] Furthermore, the defect detection result obtaining module 15 is configured to perform the following steps: extracting a first corrected crack defect feature vector and a second corrected crack defect feature vector from the updated defect memory library; Taking the crack position as an association constraint, performing association verification using the second corrected crack defect feature vector and the first corrected crack defect feature vector; if the verification fails, removing the first corrected crack defect feature vector from the updated defect memory library to obtain a stage-cleaned updated defect memory library; extracting a third corrected crack defect feature vector from the cleaning and updating defect memory bank of the stage again and performing correlation verification on the second corrected crack defect feature vector until reaching the last corrected crack defect feature vector of the cleaning and updating defect memory bank of the stage, thereby obtaining a target cleaning and updating defect memory bank; The target cleaning and updating defect memory library is identified by using a detection result identifier to obtain the defect detection result of the flexible PCB to be tested.

[0039] Furthermore, the defect detection result obtaining module 15 is further configured to perform the following steps: Extracting the first crack position center coordinates, the first crack contour area, the second crack position center coordinates, and the second crack contour area of ​​the first corrected crack defect feature vector and the second corrected crack defect feature vector; Calculating a center position deviation distance based on the center coordinates of the first crack position and the center coordinates of the second crack position, and determining whether the calculation result meets a preset position deviation threshold; if not, determining that the positions are inconsistent; Calculating an intersection-over-union (IoU) ratio based on the first crack contour area and the second crack contour area, and determining whether the IoU ratio satisfies a preset IoU ratio threshold; if not, determining that the contour diffusion is inconsistent; When the positions are inconsistent or the contour diffusion is inconsistent, the association verification result is failed.

[0040] In one possible embodiment, the position and morphology of adjacent or continuous crack correction vectors within the updated defect memory are verified pairwise, removing redundant features indicating abnormal physical locations or trends, thereby improving the credibility of the final flexible PCB defect detection results. The crack location center coordinates refer to the coordinates of the crack's geometric center point calculated using a crack morphology segmentation or location algorithm and are used for subsequent center point deviation analysis. The crack contour area refers to the enclosed region that describes the crack morphology range and can be represented by a rectangular bounding box, convex hull, or pixel set. It is primarily used to calculate the intersection over union (IOU) to determine whether cracks spread uniformly. The center position deviation distance refers to the Euclidean distance between two crack center points in space and is used to constrain the spatial consistency of the crack locations. The intersection over union (IOU) is the ratio of the intersection and union areas of two crack contour areas and is used to measure the consistency and spatial overlap of two crack morphologies.

[0041] First, the adjacent first and second corrected crack defect feature vectors are extracted from the updated defect memory library in a time series. Position and contour association verification is then performed based on this pair of vectors. Specifically, the crack position center coordinates of the two vectors are extracted, the center position deviation distance is calculated, and compared with a preset position deviation threshold set by a person skilled in the art. If the deviation is too large, the requirement is not met, the crack position is determined to be inconsistent, and the verification is marked as failed. In this case, it indicates that the first corrected crack defect feature vector may be a falsely detected crack defect due to accidental error or noise interference.

[0042] At the same time, the crack contour areas of the first corrected crack defect feature vector and the second corrected crack defect feature vector are further extracted, and the intersection-over-union (IOU) is calculated. Preferably, the ratio of the intersection to the union of the crack contour areas of the first corrected crack defect feature vector and the second corrected crack defect feature vector is calculated to obtain the IOU. If the IOU is lower than the preset IOU threshold value pre-set by a person skilled in the art, it is determined that the crack morphology diffusion is inconsistent and the verification is deemed to have failed. Once a pair of verifications fails, the previous crack correction vector is immediately removed from the updated defect memory library to form a staged cleaning and updated defect memory library.

[0043] Subsequently, the module extracts the next pair of correction vectors (such as the second and third vectors) from the cleaned library, repeating the position-contour association verification and elimination operations until the last correction vector in the library is reached, completing multiple rounds of cleaning. Finally, the module calls the detection result identifier to perform pattern clustering or confidence analysis on the valid correction vectors in the target cleaned and updated defect memory library, outputting the final defect detection results for the flexible PCB, including the presence of cracks, crack location and evolution trend, and the detection reliability score.

[0044] Preferably, a plurality of sample target cleaned and updated defect memory banks and a plurality of sample flexible PCB defect detection results to be tested are obtained as a training data set, the plurality of sample flexible PCB defect detection results to be tested in the training data set are identified, and the plurality of sample target cleaned and updated defect memory banks are input into a framework constructed based on a feedforward neural network for training, thereby obtaining output of a plurality of verification flexible PCB defect detection results to be tested. The plurality of verification flexible PCB defect detection results to be tested are compared with the plurality of identified sample flexible PCB defect detection results to be tested, and the comparison success rate is calculated. If the comparison success rate is greater than or equal to a value preset by a person skilled in the art, the verification is passed, and the trained detection result identifier is obtained.

[0045] Through multi-stage spatial and morphological consistency verification and dynamic elimination, a physically self-consistent, clean and traceable defect feature vector set is constructed, so that the detection results have high robustness and low error rate under dynamic loads.

[0046] In summary, the embodiments of the present application have at least the following technical effects: 1. This application conducts fatigue tests based on preset bending load cycle parameters, combined with multimodal synchronous acquisition of industrial cameras, flexible stress sensors and deformation sensors, to obtain video image frame sequences, stress distribution map sequences and deformation cloud map sequences covering the complete bending cycle. Compared with traditional single vision or single-point measurement, it significantly improves the real-time detection capability of hidden crack growth and local anomalies, and effectively compensates for the problem that existing static detection methods are difficult to track dynamic fatigue cracks.

[0047] 2. This application fully superimposes the flexible stress and deformation evolution information in the crack feature vector through crack time series iterative trend analysis and multi-modal physical trend weighted correction, greatly suppressing the noise and false signals in the pure visual trend. At the same time, by continuously writing the corrected crack defect feature vector into the defect memory bank, a crack evolution sample library that can be repeatedly retrieved and dynamically updated is formed, providing physically consistent, historically consistent and traceable data support for subsequent detection.

[0048] 3. This application continuously verifies and updates the crack correction feature vectors in the memory bank through repeated bending fatigue tests on the flexible PCB to be tested, and combines crack position constraints with contour intersection-over-union (IOU) verification to eliminate inconsistent or noisy samples in multiple stages, significantly reducing the cumulative error and the risk of single detection failure.

[0049] Example 2, based on the same inventive concept as the machine learning PCB defect automatic detection system in the previous embodiment, as shown in the attached Figure 2As shown, the present application provides a machine learning-based automatic PCB defect detection method. The system and method embodiments in the present application are based on the same inventive concept. The method includes: Based on preset bending load cycle parameters, a cyclic bending load is applied to the flexible PCB under test using a bending fatigue testing machine. An industrial camera, a flexible stress sensor, and a deformation sensor are used to collect data on the bending process when the first bending load is applied, thereby obtaining a first PCB bending video image frame sequence, a first flexible stress distribution map sequence, and a first deformation cloud map sequence. Performing a crack time sequence iterative trend analysis on the first PCB bending video image frame sequence to obtain a first crack implicit trend vector; performing crack defect correction on the first crack implicit trend vector according to the first flexible stress distribution map sequence and the first deformation cloud map sequence to obtain a first corrected crack defect feature vector, and storing the first corrected crack defect feature vector in a defect memory bank; performing a bending crack defect test on the flexible PCB to be tested again based on the preset bending load cycle parameters, and updating the defect memory bank according to the test results to obtain an updated defect memory bank; The updated defect memory library is integrated and associated with verification to obtain defect detection results of the flexible PCB to be tested.

[0050] Furthermore, the bending load cycle parameter is selected from at least one of the following parameters: bending radius, bending frequency, bending angle, number of cycles, and load amplitude.

[0051] Furthermore, the method further comprises: Performing trend feature analysis on the first PCB bending video image frame sequence according to the crack multi-time series bandwidth set, and sorting the frames in ascending order according to the corresponding crack multi-time series bandwidths to obtain a crack multi-time series trend vector sequence; Extracting a first crack multi-time series trend vector and a second crack multi-time series trend vector from the crack multi-time series trend vector sequence to perform implicit trend enhancement analysis and construct a first implicit trend enhancement matrix; Performing crack time series iterative trend enhancement on the second crack implicit trend vector using the first implicit trend enhancement matrix to obtain a first-stage crack implicit trend vector; Based on the first-stage crack implicit trend vector, crack time series iterative trend enhancement is performed on the third crack multi-time series trend vector in the crack multi-time series trend vector sequence, and so on to obtain the first crack implicit trend vector.

[0052] Furthermore, the method further comprises: Based on the specifications and models of the flexible PCB to be tested, historical crack development data mining is performed to obtain a historical crack development data set; Dividing the historical crack development data set according to crack type, and extracting crack development time from the division results to obtain M crack development time sets, where M is a positive integer; The maximum value and the minimum value in the M crack development time sets are respectively extracted to construct the crack multi-time series bandwidth set.

[0053] Furthermore, the method further comprises: Traversing the first flexible stress distribution map sequence and the first deformation cloud map sequence to perform feature recognition, and determining a first flexible stress feature vector sequence and a first deformation cloud map feature vector sequence; performing trend analysis based on the first flexible stress characteristic vector sequence and the first deformation cloud map characteristic vector sequence to determine a first flexible stress trend vector and a first deformation cloud map trend vector; The first crack implicit trend vector is weightedly corrected based on the first flexible stress trend vector and the first deformation cloud map trend vector to obtain a first corrected crack defect feature vector.

[0054] Furthermore, the method further comprises: extracting a first corrected crack defect feature vector and a second corrected crack defect feature vector from the updated defect memory library; Taking the crack position as an association constraint, performing association verification using the second corrected crack defect feature vector and the first corrected crack defect feature vector; if the verification fails, removing the first corrected crack defect feature vector from the updated defect memory library to obtain a stage-cleaned updated defect memory library; extracting a third corrected crack defect feature vector from the cleaning and updating defect memory bank of the stage again and performing correlation verification on the second corrected crack defect feature vector until reaching the last corrected crack defect feature vector of the cleaning and updating defect memory bank of the stage, thereby obtaining a target cleaning and updating defect memory bank; The target cleaning and updating defect memory library is identified by using a detection result identifier to obtain the defect detection result of the flexible PCB to be tested.

[0055] Furthermore, the method further comprises: Extracting the first crack position center coordinates, the first crack contour area, the second crack position center coordinates, and the second crack contour area of ​​the first corrected crack defect feature vector and the second corrected crack defect feature vector; Calculating a center position deviation distance based on the center coordinates of the first crack position and the center coordinates of the second crack position, and determining whether the calculation result meets a preset position deviation threshold; if not, determining that the positions are inconsistent; Calculating an intersection-over-union (IoU) ratio based on the first crack contour area and the second crack contour area, and determining whether the IoU ratio satisfies a preset IoU ratio threshold; if not, determining that the contour diffusion is inconsistent; When the positions are inconsistent or the contour diffusion is inconsistent, the association verification result is failed.

[0056] It should be noted that the order in which the embodiments of the present application are presented is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the foregoing descriptions of specific embodiments of this specification are provided. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential sequence shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0057] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

[0058] This specification and drawings are merely illustrative of the present application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Obviously, those skilled in the art may make various modifications and variations to this application without departing from the scope of this application. Thus, this application is intended to include such modifications and variations as fall within the scope of this application and its equivalents.

Claims

1. A machine learning PCB defect automatic detection system, characterized by: The system comprises: A deformation cloud map sequence acquisition module is used to apply a cyclic bending load to the flexible PCB under test using a bending fatigue testing machine based on preset bending load cycle parameters, and to collect data from the bending process when the bending load is first applied using an industrial camera, a flexible stress sensor, and a deformation sensor to obtain a first PCB bending video image frame sequence, a first flexible stress distribution map sequence, and a first deformation cloud map sequence; An implicit trend vector obtaining module, configured to perform crack time sequence iterative trend analysis on the first PCB bending video image frame sequence to obtain a first crack implicit trend vector; a defect feature vector storage module, configured to perform crack defect correction on the first crack implicit trend vector according to the first flexible stress distribution map sequence and the first deformation cloud map sequence, obtain a first corrected crack defect feature vector, and store the first corrected crack defect feature vector in a defect memory bank; an updated defect memory library obtaining module, configured to perform a bending crack defect test on the flexible PCB to be tested again based on the preset bending load cycle parameters, and update the defect memory library according to the test results to obtain an updated defect memory library; The defect detection result acquisition module is used to perform integrated correlation verification on the updated defect memory library to obtain defect detection results of the flexible PCB to be tested.

2. The machine learning PCB defect automatic detection system according to claim 1, characterized in that: The bending load cycle parameter is selected from at least one of the following parameters: bending radius, bending frequency, bending angle, number of cycles, and load amplitude.

3. The machine learning PCB defect automatic detection system according to claim 1, characterized in that: The implicit trend vector acquisition module is used to perform the following steps: Performing trend feature analysis on the first PCB bending video image frame sequence according to the crack multi-time series bandwidth set, and sorting the frames in ascending order according to the corresponding crack multi-time series bandwidths to obtain a crack multi-time series trend vector sequence; Extracting a first crack multi-time series trend vector and a second crack multi-time series trend vector from the crack multi-time series trend vector sequence to perform implicit trend enhancement analysis and construct a first implicit trend enhancement matrix; Performing crack time series iterative trend enhancement on the second crack implicit trend vector using the first implicit trend enhancement matrix to obtain a first-stage crack implicit trend vector; Based on the first-stage crack implicit trend vector, crack time series iterative trend enhancement is performed on the third crack multi-time series trend vector in the crack multi-time series trend vector sequence, and so on to obtain the first crack implicit trend vector.

4. The machine learning PCB defect automatic detection system according to claim 3, characterized in that: The implicit trend vector obtaining module is further configured to perform the following steps: Based on the specifications and models of the flexible PCB to be tested, historical crack development data mining is performed to obtain a historical crack development data set; Dividing the historical crack development data set according to crack type, and extracting crack development time from the division results to obtain M crack development time sets, where M is a positive integer; The maximum value and the minimum value in the M crack development time sets are respectively extracted to construct the crack multi-time series bandwidth set.

5. The machine learning PCB defect automatic detection system according to claim 1, characterized in that: The defect feature vector storage module is used to perform the following steps: Traversing the first flexible stress distribution map sequence and the first deformation cloud map sequence to perform feature recognition, and determining a first flexible stress feature vector sequence and a first deformation cloud map feature vector sequence; performing trend analysis based on the first flexible stress characteristic vector sequence and the first deformation cloud map characteristic vector sequence to determine a first flexible stress trend vector and a first deformation cloud map trend vector; The first crack implicit trend vector is weightedly corrected based on the first flexible stress trend vector and the first deformation cloud map trend vector to obtain a first corrected crack defect feature vector.

6. The machine learning PCB defect automatic detection system according to claim 1, characterized in that: The defect detection result acquisition module is used to perform the following steps: extracting a first corrected crack defect feature vector and a second corrected crack defect feature vector from the updated defect memory library; Taking the crack position as an association constraint, performing association verification using the second corrected crack defect feature vector and the first corrected crack defect feature vector; if the verification fails, removing the first corrected crack defect feature vector from the updated defect memory library to obtain a stage-cleaned updated defect memory library; extracting a third corrected crack defect feature vector from the cleaning and updating defect memory bank of the stage again and performing correlation verification on the second corrected crack defect feature vector until reaching the last corrected crack defect feature vector of the cleaning and updating defect memory bank of the stage, thereby obtaining a target cleaning and updating defect memory bank; The target cleaning and updating defect memory library is identified by using a detection result identifier to obtain the defect detection result of the flexible PCB to be tested.

7. The machine learning PCB defect automatic detection system according to claim 6, characterized in that: The defect detection result acquisition module is further configured to perform the following steps: Extracting the first crack position center coordinates, the first crack contour area, the second crack position center coordinates, and the second crack contour area of ​​the first corrected crack defect feature vector and the second corrected crack defect feature vector; Calculating a center position deviation distance based on the center coordinates of the first crack position and the center coordinates of the second crack position, and determining whether the calculation result meets a preset position deviation threshold; if not, determining that the positions are inconsistent; Calculating an intersection-over-union (IoU) ratio based on the first crack contour area and the second crack contour area, and determining whether the IoU ratio satisfies a preset IoU ratio threshold; if not, determining that the contour diffusion is inconsistent; When the positions are inconsistent or the contour diffusion is inconsistent, the association verification result is failed.

8. A machine learning-based automatic PCB defect detection method, characterized in that: The method is implemented by a machine learning PCB defect automatic detection system according to any one of claims 1 to 7, and the method comprises: Based on preset bending load cycle parameters, a cyclic bending load is applied to the flexible PCB under test using a bending fatigue testing machine. An industrial camera, a flexible stress sensor, and a deformation sensor are used to collect data on the bending process when the first bending load is applied, thereby obtaining a first PCB bending video image frame sequence, a first flexible stress distribution map sequence, and a first deformation cloud map sequence. Performing a crack time sequence iterative trend analysis on the first PCB bending video image frame sequence to obtain a first crack implicit trend vector; performing crack defect correction on the first crack implicit trend vector according to the first flexible stress distribution map sequence and the first deformation cloud map sequence to obtain a first corrected crack defect feature vector, and storing the first corrected crack defect feature vector in a defect memory bank; performing a bending crack defect test on the flexible PCB to be tested again based on the preset bending load cycle parameters, and updating the defect memory bank according to the test results to obtain an updated defect memory bank; The updated defect memory library is integrated and associated with verification to obtain defect detection results of the flexible PCB to be tested.

Images