A method for detecting correlation between front and back defects of SMT patches

Abstract

This invention relates to the field of printed circuit board assembly and inspection technology, and discloses a method for detecting defects before and after SMT placement. The method first establishes a unified reference system that spans solder paste inspection, placement, pre-reflow optical inspection, and post-reflow optical inspection by reading the board-level barcode and identifying reference markers on the board surface. It then converts the inspection data from each workstation to this unified reference system and establishes physical target identification based on the board-level barcode, pad number, component affiliation, package type, and relative orientation of pad clusters. Next, it aggregates the same solder joint or component before and after reflow across processes to form a process sequence evidence chain. The reflow self-alignment convergence relationship is verified by combining the post-reflow wetting state and solder joint forming state, and confidence correction is applied to the single-workstation defect judgment. Confirmed defects are traced back to the first deviation workstation, and parameter compensation information is generated. This method effectively reduces post-reflow misjudgments and missed defects, improves the consistency of defect judgment, and enhances root cause tracing capabilities.

Description

Technical Field

[0001] This invention relates to the field of printed circuit board assembly and inspection technology, specifically to a method for detecting defects before and after SMT assembly. Background Technology

[0002] Surface mount technology (SMT) is a commonly used manufacturing process in the electronics assembly field. It typically includes solder paste printing, solder paste inspection (SPI), component placement, pre-reflow automated optical inspection (AOI), reflow soldering, and post-reflow automated optical inspection (AOI). The inspection results of each step directly affect the accuracy of defect judgment and process yield of printed circuit board assemblies. For example, the published invention patent application CN108960306B discloses a solder paste inspection threshold optimization method based on SMT big data, which optimizes the solder paste inspection threshold by correlating SPI inspection data, AOI inspection data, and repair data. Another example is the published invention patent application CN113094980B, which discloses a solder paste printing quality prediction method and system based on IGA-DNN, which predicts the solder paste printing quality based on solder paste printing process parameters.

[0003] While the aforementioned technologies can optimize solder paste detection thresholds or printing quality, they primarily rely on single-process detection results or offline statistical data for judgment. They fail to establish a continuous characteristic correspondence and causal evidence chain between the same physical solder joint or component in SPI, pre-reflow AOI, and post-reflow AOI. In actual production, the reflow soldering process exhibits a solder self-alignment effect. Mounting misalignment and solder paste quantity deviations detected in upstream processes may converge spontaneously after reflow, while solder joints deemed acceptable upstream may develop into actual defects after reflow due to insufficient wetting, bridging, or uncorrected solder joint position deviations. Furthermore, existing detection methods typically rely on static images acquired independently at a single moment from each workstation. The lack of unified analysis of the causal relationship between upstream deviations, reflow self-healing processes, and post-reflow forming states among workstations leads to self-healing process deviations being easily misjudged as defects, while real defects formed by deterioration after reflow may be missed. Furthermore, it is difficult to reverse-locate post-reflow defects to specific root cause processes such as printing, mounting, or reflow, thereby increasing the workload of manual review and reducing the consistency of defect judgment and process traceability efficiency. Therefore, there is an urgent need to propose a detection method that can perform cross-process feature registration, evidence chain construction, defect confidence correction, and process root cause attribution around the same physical solder joint or the same physical component, in order to solve the problems of misjudgment in post-reflow processes, missed detection of real defects, and difficulty in tracing the root cause of defects. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method for detecting defects before and after SMT assembly, which solves the problems mentioned in the background section.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A method for detecting defects before and after SMT assembly includes:

[0007] S1: Read the printed circuit board level barcode, identify the reference mark points on the board surface, and establish a unified reference system that runs through solder paste inspection, mounting, pre-reflow optical inspection and post-reflow optical inspection;

[0008] S2: Convert the pad positions, component positions, solder paste status, mounting positions, pre-reflow overlap status and post-reflow forming status collected from each workstation to a unified reference system, and establish physical target identification based on board-level barcode, pad number, component affiliation, package type and relative orientation of pad clusters.

[0009] S3: Based on the physical target identification, collect the same solder joint or the same component before and after reflow, and attach the status information and deviation information of each station according to the process sequence;

[0010] S4: Based on upstream deviation information, package type, pad size, solder paste wetting state and post-reflow forming state, verify the reflow self-alignment convergence relationship and correct the single-station defect judgment confidence state.

[0011] S5: For defects confirmed after confidence correction, trace back upstream along the process sequence to determine the first deviation from the workstation, and feed back the direction and amount of deviation to the corresponding process for parameter compensation.

[0012] Furthermore, S1 includes:

[0013] Collect board-level barcodes, board surface images, nominal coordinates of reference markers, and measured coordinates of reference markers. When barcode reading fails, generate a temporary board identifier based on the vehicle number and transit time.

[0014] Brightness equalization and edge enhancement are performed on the board image. The contour features of three non-collinear global reference markers and local reference markers are identified. Two-dimensional similarity transformation or affine transformation relationship is established based on the same reference markers. Registration residuals are calculated by region, and reference system number, coordinate conversion relationship, residual record, valid status and anomaly mark are generated.

[0015] When the identification of the reference marker fails, the image is resampled or a backup local reference marker is enabled. When the local residual exceeds the limit, a low confidence spatial state is set.

[0016] Furthermore, S2 includes:

[0017] Based on a unified reference system number and coordinate conversion relationship, coordinate transformation, time recording standardization, and board orientation correction are performed on the data of each workstation.

[0018] Perform mirror calibration on the double-sided panel and correct the station overtime record based on the first code reading time and the production line cycle window.

[0019] Based on board-level barcode, component tag number, pad number, package type, pad size, nominal center position of pad, and relative distance, relative direction and arrangement order within the pad cluster, a physical target identity record is established, binding the solder paste state, mounting position, pre-reflow overlap state and post-reflow forming state to the corresponding pad or component.

[0020] When names are inconsistent, tag number mapping is performed; when fields are missing, encapsulation matching, identity confidence marking, and identity anomaly recording are performed.

[0021] Furthermore, mirror calibration is performed on the double-sided panels, and the overtime records are corrected using the first code reading time and production line cycle time window, including:

[0022] Establish board orientation marks for the front and back sides respectively, and perform mirror conversion of the back coordinates based on the flip-board relationship;

[0023] The timing anchor point is the first time the solder paste is read at the inspection station. The production line cycle window is determined based on the continuous station pass records, conveyor speed and board spacing. The window width is less than half of the minimum time interval between adjacent boards. The subsequent station pass records are then corrected according to the production line cycle window.

[0024] Update window parameters when production line speed or board spacing changes.

[0025] Furthermore, S3 includes:

[0026] Read the target record of the same board based on the physical target identity table and the workstation data standardization table;

[0027] Before reflow, the components are associated with the pad number, component affiliation and relative orientation of the pad cluster. After reflow, the overall matching is performed according to the relative distance, relative direction, arrangement order, adjacent spacing ratio and endpoint orientation within the solder joint group.

[0028] Auxiliary aggregation of BGA arrays is performed based on corner orientation, edge rows and columns, and adjacent complete rows and columns;

[0029] Establish a chain of evidence based on printing process, solder paste inspection, mounting, pre-reflow inspection, reflow process, and post-reflow inspection, and record the collection confidence status, data breakpoints, barcode mixed reading correction results, and collection anomalies.

[0030] Furthermore, the BGA array is further aggregated by corner orientation, edge rows and columns, and adjacent complete rows and columns, including:

[0031] For BGA solder joint arrays, when some solder joints are not effectively extracted due to obstruction, reflection, insufficient solder, or recognition failure, the orientation of the identified corner points, the direction of the edge rows and columns, and the spacing between adjacent complete rows and columns are read.

[0032] After confirming the location of a corner point or the row and column directions of two mutually perpendicular edges, array assignment matching is performed.

[0033] If the corner point orientation and edge row / column direction are not confirmed, or if the relative relationship of solder joint groups does not match the relative relationship of nominal pad clusters, it is recorded as a BGA array aggregation anomaly.

[0034] Furthermore, S4 includes:

[0035] Determine the tolerable deviation margins for solder paste quantity, placement offset, rotation angle, and package type based on the cross-process evidence chain;

[0036] The self-alignment status is verified by the offset direction in front of the furnace, the displacement direction behind the furnace, and the change in offset behind the furnace.

[0037] By combining the wetting boundary after the furnace, solder joint height, solder continuity, solder balls, tombstoning, and aggregated confidence status, the conclusions of single-station defects are jointly judged, confidence marked, downgraded, and image re-sampling or manually reviewed.

[0038] Furthermore, the self-alignment status is verified according to the offset direction in front of the furnace, the displacement direction behind the furnace, and the change in offset behind the furnace, including:

[0039] For placement offset targets, read the offset direction and offset magnitude of the placement node or the pre-reflow detection node, and take the direction from the offset position to the center of the pad as the expected convergence direction;

[0040] Read the final position of the detection node after the furnace, determine the displacement direction and offset amplitude after the furnace, calculate the difference between the offset amplitude before the furnace and the offset amplitude after the furnace, and perform self-alignment status verification based on the angle between the expected convergence direction and the displacement direction after the furnace, the change in offset amplitude, and the abnormal markers of insufficient wetting, bridging, and tombstoning.

[0041] Furthermore, S5 includes:

[0042] Based on the final defect determination results, the cross-process evidence chain, and the deviation information of each node, the reflow process node, the pre-reflow process node, the mounting node, the solder paste inspection node, and the printing process node are traced back sequentially from the post-reflow inspection node.

[0043] The main causative process and cooperating factors are determined based on the correspondence between deviation and post-furnace defect type, and compensation information for printing, mounting or reflow processes is generated respectively.

[0044] When the root cause is not unique, a list of root cause candidates is output. After manual confirmation, parameter adjustments are performed, the occurrence rate of similar defects is tracked, process drift warnings are generated, and invalid compensation records and parameter rollbacks are recorded.

[0045] Furthermore, based on the correspondence between deviations and post-reflow defect types, the main causative processes and synergistic factors are determined, and compensation information for printing, mounting, or reflow processes is generated, including:

[0046] Corresponding solder paste volume deviation and solder paste deposition offset to insufficient solder, cold solder joint or bridging type, generating printing adjustment information, the printing adjustment information records pad area, deviation ratio, deviation direction, stencil cleaning prompt, squeegee pressure adjustment direction and printing speed adjustment direction.

[0047] The landing point offset and rotation angle are correlated with the continuous offset or tombstone type to generate placement compensation information. The placement compensation information records the nozzle number, component number, average offset vector and landing point compensation direction.

[0048] By correlating peak temperature, time above the liquidus line, and heating rate with the types of poor wetting, solder balls, or solder overflow, reflow adjustment information is generated.

[0049] When multiple workstations deviate, the primary cause process is determined by the workstation that deviates first, and subsequent workstations that deviate are marked as contributing factors.

[0050] Compared with the prior art, the present invention provides a method for detecting defects before and after SMT assembly, which has the following beneficial effects:

[0051] 1. This invention, through board-level barcodes, reference markers, and a unified reference system, aggregates data from solder paste inspection, mounting, pre-reflow optical inspection, and post-reflow optical inspection to the same physical solder joint or the same physical component. It also establishes a cross-process evidence chain by combining the relative orientation of solder pad clusters, pre-reflow deviation information, post-reflow wetting state, and solder joint forming state. Based on this, the self-alignment convergence relationship during reflow soldering is verified, and the confidence correction of single-station defect conclusions is performed according to the cross-process evidence chain. The confirmed defects are then traced back along the process timeline to the first deviation station, generating parameter compensation information for the corresponding process. This achieves the effect of reducing excessive post-reflow scrapping caused by self-healing process deviations and missing true defects after reflow, improving the consistency of defect judgment, the efficiency of manual review, and the ability to trace the root cause of processes.

[0052] 2. This invention achieves unified coordinate transformation, unified time recording, board orientation correction, and physical target identity binding for inspection data from various workstations. It also performs low-confidence marking, anomaly recording, image re-acquisition, and manual review when barcode reading anomalies, reference point recognition failures, field missingness, partial missing BGA solder joints, or evidence chain breaks occur. This ensures that inspection data from different equipment, different boards, and different workstations can be collected, verified, and transmitted in a consistent data format. Simultaneously, deviation information from printing, mounting, and reflow processes is written into batch-level process records, and process drift warnings, compensation verification, and parameter rollback are performed based on similar defect occurrences. This improves cross-equipment data consistency, anomaly data processing stability, and batch-level process control capabilities. Attached Figure Description

[0053] Figure 1 This is a schematic diagram of the process for detecting defects before and after SMT assembly according to the present invention;

[0054] Figure 2 This is a schematic diagram illustrating the establishment of a unified reference system and the multi-station coordinate transformation of the present invention;

[0055] Figure 3 This is a schematic diagram illustrating the association between the physical target identity identifier and the pad cluster in this invention;

[0056] Figure 4 This is a schematic diagram illustrating the construction of a cross-process evidence chain in this invention;

[0057] Figure 5 This is a logic diagram of the backflow self-alignment convergence relationship and confidence correction of the present invention;

[0058] Figure 6 This is the closed-loop diagram of root cause tracing and parameter compensation in this invention. Detailed Implementation

[0059] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] Example 1: Figure 1 - Figure 6 A method for detecting defects before and after SMT assembly is presented, including:

[0061] S1: Read the printed circuit board level barcode, identify the reference mark points on the board surface, and establish a unified reference system that runs through solder paste inspection, mounting, pre-reflow optical inspection and post-reflow optical inspection;

[0062] S2: Convert the pad positions, component positions, solder paste status, mounting positions, pre-reflow overlap status and post-reflow forming status collected from each workstation to a unified reference system, and establish physical target identification based on board-level barcode, pad number, component affiliation, package type and relative orientation of pad clusters.

[0063] S3: Based on the physical target identification, collect the same solder joint or the same component before and after reflow, and attach the status information and deviation information of each station according to the process sequence;

[0064] S4: Based on upstream deviation information, package type, pad size, solder paste wetting state and post-reflow forming state, verify the reflow self-alignment convergence relationship and correct the single-station defect judgment confidence state.

[0065] S5: For defects confirmed after confidence correction, trace back upstream along the process sequence to determine the first deviation from the workstation, and feed back the direction and amount of deviation to the corresponding process for parameter compensation.

[0066] This method is applied to the surface mount production process of printed circuit board (PCB) assemblies and is applicable to the manufacturing and testing of PCB components. The production process sequentially follows the flow: solder paste printing, solder paste inspection (SPI), component mounting, pre-reflow optical inspection, reflow soldering, and post-reflow optical inspection. The processed objects include at least PCBs with board-level barcodes, board reference markers, pads, components, solder paste deposition areas, pre-reflow solder joint overlap areas, and post-reflow solder joint forming areas. Input information includes at least PCB design data, solder paste inspection data, mounting equipment records, pre-reflow optical inspection data, post-reflow optical inspection data, and reflow process records. The PCB design data includes at least the nominal coordinates of the pads, component reference designators, package type, pad size, component orientation, and reference marker position. The solder paste inspection data includes at least the solder paste volume, solder paste area, solder paste height, and solder paste deposition rate. The data should include fields such as center offset, component placement equipment records, and at least the component placement coordinates, rotation angle, and placement time. Pre-reflow optical inspection data should include fields such as the overlap state between the solder joint and the pad, pre-reflow offset direction, and pre-reflow offset magnitude. Post-reflow optical inspection data should include fields such as solder joint spread shape, wetting state, bridging state, forming anomaly type, and final position. During processing, the same physical solder joint or component is used as the object. Inspection information collected at different stations and times is aggregated into the same reference frame. The convergence state of upstream deviations after reflow is verified based on the reflow soldering self-alignment law. Confidence correction is applied to the single-station inspection conclusions. Confirmed defects are then traced back along the process sequence to the first deviation station. Output data should include at least the joint judgment result, confidence state, root cause process, deviation direction, deviation amount, and parameter compensation information for each physical target.

[0067] Specifically, such as Figure 2 As shown: When a printed circuit board enters the solder paste inspection station or the first inspection station with image acquisition capabilities, the board-level barcode is read and a unified reference system establishment process is triggered. The information read includes at least the board-level barcode, board image, nominal coordinates of reference markers in the design data, and measured coordinates of reference markers in the current station image. The board-level barcode is used to identify the current individual printed circuit board, and the reference markers are used to establish the correspondence between the inspection image coordinates and the design coordinates. The board-level barcode can be obtained by recognizing QR codes, barcodes, or laser-engraved characters. When the board-level barcode reading fails, a temporary board identifier is generated based on the production line carrier number and upstream transit time. The temporary board identifier is only used for data caching, re-acquisition and positioning, and anomaly tracking, and is not used as the basis for board identification in defect judgment.

[0068] The reference markers can be selected as at least three non-collinear global reference markers to establish correction relationships for translation, rotation, scaling, and local tilt errors of the board surface, and to identify abnormal points through the residual differences of each reference marker. In double-sided mixed-assembly boards, irregularly shaped boards, BGA areas, or fine-pitch package areas, local reference markers close to the area to be detected are further read as the basis for regional correction. When identifying reference markers, the board surface image is first subjected to brightness equalization and edge enhancement, and then the reference center coordinates are extracted based on the circular, cross-shaped, or rectangular contour features. When there is reflection, solder paste contamination, or local occlusion in the image, abnormal reference markers are removed, and the valid reference markers that are not occluded are used to participate in the subsequent establishment of conversion relationships.

[0069] After obtaining the nominal and measured coordinates of the reference markers, a conversion relationship from the current workstation coordinates to the unified reference system is established based on the positional correspondence between the corresponding reference markers. For planar panels, a two-dimensional similarity transformation is used for translation, rotation, and scaling correction. When there are slight perspective deviations or local imaging distortions in the panel image, an affine transformation is used for correction. After the conversion relationship is established, each reference marker is transformed from the workstation coordinates to the unified reference system, and the transformed reference center coordinates are compared with the design nominal coordinates to obtain the reference point registration residual. The coordinate deviation of each reference marker is the planar distance between the transformed reference center coordinates and the design nominal coordinates. The registration residual can be selected as the root mean square value obtained by averaging the squares of the coordinate deviations of all valid reference markers. In fine-pitch areas or areas with densely packed adjacent pads, the registration residual can also be selected as the maximum coordinate deviation to improve the sensitivity of abnormal offset identification.

[0070] For standard pitch pad areas, the reference point registration residual should be within 30 micrometers; for BGA peripheral pads, fine-pitch packages, or areas with a pad center distance of less than 0.5 millimeters, the reference point registration residual should be within 20 micrometers. 30 micrometers and 20 micrometers are preferred control values, which can be determined by multiple data acquisitions and calibrations on the same production line using a standard calibration board or a first-piece qualified board, combined with the repeatability of SPI and AOI equipment, pad center distance, and the distinguishable distance between adjacent pads. Standard pitch pad areas have relatively large space margins, and 30 micrometers can cover normal positioning fluctuations of the equipment and reduce the risk of cross-station mismatch. A pad center distance of less than 0.5 millimeters corresponds to the process scale of fine-pitch packages commonly used in surface mount technology. The distinguishable distance between adjacent pads in this area is even smaller, so the registration residual is tightened to 20 micrometers to avoid data from the pre-reflow and post-reflow stages being linked to adjacent pads.

[0071] The unified reference system data includes at least the unified reference system number of the current printed circuit board, the conversion relationship between the coordinates of each workstation and the unified reference system, the residual record of the reference marker point, the valid status of the reference system, and the abnormal marker, etc., and is transmitted to each workstation for data conversion and physical target identification establishment process. If the reference marker point of any workstation fails to be identified, the workstation is triggered to re-acquire the board image. If it still cannot be identified after re-acquisition, the backup local reference marker point is read. If the backup local reference marker point still cannot meet the residual requirements, the workstation is recorded as a reference system abnormal workstation, and the data of this workstation is not used as a basis for high confidence judgment. If only the residual of a local area exceeds the limit, while the residual of other areas meets the residual requirements, the pads, components, solder paste deposition areas and solder joint forming areas in the area exceeding the limit are set to a low confidence space state and used as auxiliary evidence in subsequent cross-process aggregation.

[0072] Specifically, such as Figure 3 As shown: After generating a unified reference system number and coordinate conversion relationship, coordinate transformation and standardized recording format are applied to the data collected from each workstation. The input data includes at least the following fields: coordinate conversion relationship, printed circuit board design data, solder paste inspection data, placement equipment records, pre-reflow optical inspection data, and post-reflow optical inspection data. The pad positions, component positions, solder paste states, placement positions, pre-reflow overlap states, and post-reflow forming states collected from each workstation are converted to a unified reference system and recorded using a unified coordinate unit, a unified time recording method, and a unified board orientation. The coordinate unit can be selected as micrometers to adapt to the micrometer-level position data output by SPI and AOI devices and to maintain consistency with the coordinate conversion accuracy of the unified reference system. The time recording method can be selected as millisecond-level timestamps to record the detection acquisition time and production line transit time, and to distinguish the acquisition order of consecutive printed circuit boards at the same workstation.

[0073] For double-sided boards, board orientation markings are established for sides A and B respectively, and the coordinate orientation after flipping is mirrored to ensure that the detection data of the front and back sides have a consistent orientation under a unified reference system. For cases where the same printed circuit board repeatedly reads board-level barcodes at multiple stations, the barcode reading time at the first entry into the solder paste inspection station is used as the timing anchor point, and the transit records of subsequent stations are corrected according to the production line cycle time window. The production line cycle time window is determined based on the minimum time interval between two adjacent printed circuit boards passing through the same inspection station. This minimum time interval can be obtained statistically from continuous transit records or calculated based on the conveyor line speed and the spacing between adjacent boards. The window width is less than half of this minimum time interval, ensuring that at most one printed circuit board corresponds to the same window. When the production line speed or board spacing is adjusted, the production line cycle time window is updated.

[0074] Physical target identification uses board-level barcodes as individual board identifiers, pad reference numbers and component affiliation as basic object identifiers, and package type, pad size, nominal center position of the pad, and relative orientation of pad clusters as enhanced identifiers. When establishing physical target identification, the current printed circuit board individual is first determined based on the board-level barcode. Then, the physical component corresponding to each pad is determined based on the component reference number, pad number, and package type in the design data. All pads corresponding to the same component are grouped into the same pad cluster. Within the pad cluster, the relative distance, relative direction, and arrangement order between each pad are recorded. For example, for chip resistors and capacitors, the pad cluster... It consists of two end pads, and records the center spacing and center connection direction of the two end pads; for small outline integrated circuits, the pad cluster records the pin order on the same side and the pin correspondence on the opposite side according to the pin arrangement direction; for square flat leadless package devices, the positional relationship between the peripheral pad group and the central heat dissipation pad is recorded; for ball grid array package devices, the row and column positions and corner orientations of the ball pads are recorded; the relative orientation of the pad cluster is not based on the coordinates of a single solder point as the sole matching basis, but on the stable geometric relationship between the pads inside the same component as the identification basis, in order to adapt to the situation where the component undergoes overall translation or slight rotation after reflow;

[0075] After solder paste status data is converted to a unified reference system, it is bound to the corresponding pad identification. Solder paste volume, solder paste area, and solder paste height are converted to proportional values ​​relative to the nominal values ​​of the corresponding package type. The deviation of the solder paste deposition center relative to the geometric center of the pad is recorded according to the offset direction and offset magnitude. After the placement position data is converted, it is bound to the component identification, and the offset direction, offset magnitude, and rotation angle of the component placement center relative to the nominal center of the pad cluster are recorded. The pre-reflow overlap status is bound to the component identification, and the actual overlap ratio, offset direction, and position shape of the solder end covering the pad are recorded. The post-reflow forming status is bound to the solder joint identification, and the solder joint spreading range, wetting boundary, solder joint center offset, bridging relationship, and forming anomaly type are recorded.

[0076] The physical target identity table should include at least the following fields: board-level barcode, board orientation, component reference number, pad number, package type, pad size, nominal center position of pad, relative orientation of pad cluster, and identity confidence status. The workstation data standardization table should include at least the following fields: workstation name, acquisition time, unified reference system coordinates, status field, deviation field, data source, and data quality marker, and should be used for subsequent cross-process aggregation and evidence chain linking.

[0077] If the component reference number in the design data is inconsistent with the name output by the testing equipment, a pre-established reference number mapping table is used to unify the names. If the testing equipment only outputs the pad area number and not the component reference number, the corresponding component is looked up based on the nominal coordinates of the pad, the relative orientation of the pad cluster, and the package geometry. When the package type needs to be inferred from the number of pads, the pad arrangement, or the component's external dimensions, or when the testing equipment output fields are missing but the corresponding component can still be looked up based on the relative orientation of the pad cluster, the identity confidence status is recorded as low confidence. When the above identity fields and coordinate transformation status can be directly correlated, the identity confidence status is recorded as high confidence. When the board-level barcode, coordinate position, and relative orientation of the pad cluster cannot be correlated, the corresponding data is recorded as identity abnormal data. Low confidence can participate in subsequent cross-process aggregation, but it is not the sole basis for automatic high confidence determination. Thus, the same physical solder joint or the same physical component with inconsistent names, coordinates, and testing items in different workstations can be classified into the same identity identifier.

[0078] Specifically, such as Figure 4 As shown: After the physical target identification table and the workstation data standardization table are generated, the inspection records of the same physical solder joint or the same physical component in each workstation are collected based on the board-level barcode and the physical target identification. The input data includes at least the fields of the physical target identification table, the workstation data standardization table, the acquisition time of each workstation, the production line cycle record, and the spatial position data before and after reflow soldering. During the collection, the same printed circuit board is first identified based on the board-level barcode, and then the physical target identification is used to read the physical target's performance in solder paste inspection, mounting, pre-reflow optical inspection, and post-reflow. Corresponding records in optical inspection; For the pre-reflow station, the pad position and component position are usually close to the design nominal position, and can be associated with the pad number, component affiliation and relative orientation of the pad cluster under a unified reference system; For the post-reflow station, due to the surface tension generated after the solder melts during reflow soldering, the component may undergo self-alignment displacement towards the center of the pad. Post-reflow aggregation does not use a single post-reflow measured coordinate and fixed pad coordinate as the only matching basis, in order to avoid aggregation mismatch in fine-pitch components, densely adjacent pad areas or after the overall translation of components;

[0079] Post-reflow data collection employs a pad cluster overall relationship matching method. Using the same component solder joint group detected post-reflow as the matching target, the relative distance, relative direction, and arrangement order of each solder joint within the group are read and compared with the nominal pad cluster relative orientation recorded in the physical target identification. During comparison, the entire pad cluster is allowed to undergo translation and small-angle rotation consistent with reflow self-alignment, but the arrangement order of solder joints within the group, the adjacent spacing ratio, and the endpoint orientation should meet the matching conditions for the corresponding package type. For surface mount components, the deviation between the center line direction of the two end solder joints and the nominal direction should not exceed 5°, and the deviation between the center distance of the two end solder joints and the nominal center distance should not exceed 10%. Among these, 5... ° is used to cover AOI recognition angle errors, slight rotation caused by reflow self-alignment, and normal placement fluctuations of chip components; 10% is used to accommodate solder joint spreading boundary recognition errors, solder paste forming boundary changes, and solder joint center extraction errors; for small outline integrated circuits, the solder joint arrangement order on the pin side remains consistent, and the spacing deviation between adjacent solder joints is no more than 15%; this value is higher than the center distance deviation of chip components because the multi-pin solder joint boundaries of small outline integrated circuits are more susceptible to wetting spreading, reflection, and pin shading. Using 15% as the aggregation matching tolerance can take into account both recognition errors and array order stability; 5°, 10% and 15% are all used for cross-process aggregation matching and are not used as the criteria for judging whether solder joint defects are acceptable;

[0080] Matching tolerance can be determined by collecting the center positions of solder joints of similar packages before and after reflow using the first qualified board or standard calibration board, and by statistically analyzing the directional deviation, center distance deviation, and adjacent solder joint spacing deviation after normal reflow. For situations where the accuracy of the testing equipment is improved, the package spacing is reduced, or the quality control requirements of the production line are increased, the matching tolerance can be tightened accordingly. For situations where solder joints have strong reflectivity, wide wetting boundaries, or large image edge extraction errors, the matching tolerance should be reviewed in conjunction with manually confirmed samples, without changing the arrangement order and relative orientation of the solder pad clusters. The setting of the matching tolerance is based on the process characteristic that the arrangement order and relative position of solder joints within the same component remain basically stable after normal reflow. If the relative relationship within a solder joint group is significantly disrupted, it will not be processed according to normal aggregation.

[0081] For BGA solder joint arrays, array affiliation is determined based on corner solder joints, edge row and column directions, and the spacing between solder joints in adjacent complete rows and columns. When some solder joints are not effectively extracted due to obstruction, reflection, insufficient solder, or identification failure, auxiliary affiliation is performed using identifiable corner locations, edge row and column directions, and adjacent complete rows and columns. During auxiliary affiliation, at least one corner location or two mutually perpendicular edge row and column directions are retained. When neither corner location nor edge row and column direction can be confirmed, or when the relative relationship of solder joint groups cannot be established with the nominal pad cluster, it is recorded as a BGA array affiliation anomaly. BGA array affiliation anomalies are not forcibly linked to adjacent pads, but are used as auxiliary clues for defect determination, such as severe offset, tombstoning, bridging, or missing solder joints.

[0082] After completing the post-reflow data collection, the status and deviation information of each workstation for the same physical target are sequentially linked according to the process sequence. Cross-process evidence chain nodes are arranged according to solder paste inspection node, mounting node, pre-reflow inspection node, and post-reflow inspection node. When the printer can output squeegee pressure, printing speed, demolding speed, or stencil cleaning status, a printing process node is added before the solder paste inspection node. When the reflow oven can output peak temperature, time above the liquidus line, or heating rate, a reflow process node is added between the pre-reflow and post-reflow inspection nodes. Each node records at least the workstation name, collection time, status description, and deviation direction. Fields include deviation range, correspondence with the preset process allowable range, data quality identifier, and anomaly identifier; among which, the preset process allowable range is determined by the process rules corresponding to solder paste volume, placement offset, solder joint overlap, and post-reflow forming state; the solder paste inspection node records the solder paste volume ratio, solder paste area ratio, solder paste height ratio, and solder paste deposition center offset; the placement node records the landing point offset, angle deviation, and placement equipment station number; the pre-reflow inspection node records the solder joint overlap ratio and pre-reflow offset; the post-reflow inspection node records the solder joint forming state, wetting state, bridging relationship, low solder state, tombstoning state, offset state, and final position;

[0083] The cross-process evidence chain, based on the same physical target, records the state evolution and deviation changes according to the process sequence, for use in verifying the reflow self-alignment convergence relationship and correcting the confidence of single-station defect judgment. After collection, a cross-process evidence chain and collection confidence status are generated. The collection confidence status is used for confidence adjustment of defect judgment. When the solder paste inspection, mounting, pre-reflow inspection, and post-reflow inspection nodes are all complete, and the post-reflow solder joint group meets the overall relationship matching condition of the solder pad cluster, the collection confidence status is recorded as high confidence. When the evidence chain has a locatable breakpoint and can still form a sub-chain, the collection confidence status is recorded as medium confidence. When the post-reflow collection relies on auxiliary rows and columns, local references, or low-confidence identities, the collection confidence status is recorded as low confidence. When the board-level barcode, the relative orientation of the solder pad cluster, and the time window cannot establish a corresponding relationship, the corresponding data is recorded as an overall collection anomaly.

[0084] If data from a certain intermediate station is missing, for example, if the pre-furnace optical inspection does not collect a valid image, the breakpoint is recorded at the corresponding node, and a sub-chain is formed based on the available station records. The sub-chain continues to participate in defect judgment. If multiple printed circuit boards have barcodes read in adjacent time windows, correction is performed based on the board-level barcode scanning event, carrier position, and production line cycle time. Only when there is a unique corresponding printed circuit board within the time window will the corresponding station data be attached to that board. In this way, the pre-furnace and post-furnace inspection information are grouped into the same physical target, providing a data basis for reflow self-alignment consistency verification.

[0085] Specifically, such as Figure 5 As shown: After the cross-process evidence chain and aggregated confidence state are generated, the reflow self-alignment convergence relationship is verified based on the upstream deviation information and post-reflow forming state of the same physical target. The input data includes at least the fields of cross-process evidence chain, aggregated confidence state, solder paste deviation information, mounting deviation information, package type, pad size, pre-reflow overlap state, post-reflow wetting state, and post-reflow forming state. The verification of the reflow self-alignment convergence relationship is based on the surface tension generated by the molten solder during the reflow soldering process. When the solder tip and the pad maintain effective overlap, the mounting offset within a reasonable range can converge after reflow. When the amount of solder paste, the solder tip overlap ratio, the mounting rotation angle, or the component weight exceeds the self-alignment capability range, the deviation is difficult to completely correct through reflow self-alignment and may manifest as insufficient solder, cold solder joints, bridging, tombstoning, or continuous offset after reflow.

[0086] First, determine the tolerable deviation margin for each physical target. For solder paste condition, the solder paste volume can be selected as 50% to 150% of the nominal volume as the basic process window. The solder paste area and height, combined with the package type, pad size, and stencil opening parameters, are used to jointly determine the solder paste deposition state. This range is determined based on process control experience in SPI testing, where insufficient solder paste corresponds to risks of low solder and cold solder joints, while excessive solder paste corresponds to risks of bridging and solder balls. This range can also be calibrated through first article confirmation, standard sample verification, or historical yield data statistics. When similar packages form continuous wetting after normal reflow and no bridging occurs, the corresponding solder paste volume distribution can be used as the basis for correcting the process window for that package. When the solder paste area is significantly smaller than the pad opening coverage area, the judgment direction is biased towards low solder risk. When the solder paste height increases abnormally and the area expands outward, the judgment direction is biased towards risks of solder accumulation, solder balls, or bridging. For fine-pitch packages, high-reliability products, or areas with small pad center distances, the solder paste volume window can be narrowed after first article confirmation or process verification.

[0087] For surface mount resistors and capacitors, the placement offset can be set to no more than 50% of the pad width. This ratio ensures that the solder joints maintain approximately half of their effective overlap, providing a foundation for surface tension pullback during reflow. When the effective overlap is below this ratio, the pullback torque is insufficient, and the component risks being pushed off the pad or developing tombstoning. For rotational deviation of surface mount components, the rotation angle deviation can be set to no more than ±5°. This range is used to determine the self-alignment capability of the placement rotation deviation, which differs from the directional tolerance in the post-reflow solder joint grouping and matching. The ±5° setting is because a small rotation angle allows for better wetting by the difference in solder wetting power at both ends. The generated rotational torque is corrected, but large-angle rotation can easily cause one end of the solder pad to detach from the pad. For packages such as QFN and BGA with large bottom solder pad areas or heavy components, the upper limit of self-alignment displacement can be selected as 60% to 80% of the upper limit corresponding to the chip component. That is, it is tightened based on the upper limit of displacement determined by the chip component based on 50% of the pad width. This ratio is determined based on the process characteristics that the self-alignment capability of the bottom terminal package is weaker than that of the chip component due to the influence of component weight, bottom wetting area, heat dissipation pad constraint and solder ball collapse process during reflow. It is used to avoid mistaking the misalignment of heavy components that are difficult to self-heal as self-healable deviation.

[0088] Subsequently, the self-alignment convergence relationship is verified based on upstream deviation information and post-reflow forming status. For placement offset targets, the offset direction and magnitude in the placement node or pre-reflow detection node are read, with the direction from the offset solder end to the center of the solder pad taken as the expected convergence direction. Then, the final position of the solder joint or component in the post-reflow detection node is read, and it is compared whether the post-reflow offset magnitude is less than the pre-reflow offset magnitude, and whether the actual displacement direction has a component pointing towards the center of the solder pad. The consistency of direction can be represented by the angle between the expected convergence direction and the actual post-reflow displacement direction, which can be selected to be no greater than 45°. This angle corresponds to the actual displacement direction still having a major component pointing towards the center of the solder pad, which can accommodate equipment measurement errors and slight rotation during reflow, while also eliminating... In addition to lateral offset or displacement significantly away from the pad center, for fine-pitch packages, high-reliability products, or areas with small pad center-to-center distances, this included angle can be tightened to 30° to reduce the risk of slight lateral offset being misjudged as convergence towards the pad center. The convergence range is expressed as the difference between the pre-reflow offset range and the post-reflow offset range. When the post-reflow offset range is less than the pre-reflow offset range, the actual displacement direction meets the direction consistency requirements, and there is no insufficient wetting, bridging, or tombstoning after reflow, the upstream offset is judged to be consistent with the reflow self-alignment convergence relationship. When the post-reflow offset does not converge, the offset further expands, the actual displacement direction deviates from the pad center, or the post-reflow is accompanied by abnormal solder joint formation, the deviation is judged to be inconsistent with the reflow self-alignment convergence relationship.

[0089] For solder paste quantity deviation targets, the solder paste volume ratio, solder paste area ratio, solder paste height ratio, and solder paste deposition center offset are read from the solder paste detection nodes and verified in conjunction with the post-reflow wetting state and solder joint formation state. The post-reflow wetting state and solder joint formation state are determined by the solder joint spreading boundary, wetting continuity, solder joint height, solder outline, and solder connectivity between adjacent pads obtained from post-reflow optical inspection. When the solder paste volume is low but still within the tolerable deviation margin, and the post-reflow wetting boundary is continuous, the spreading shape is complete, and the solder joint height does not decrease abnormally, the deviation is identified. Do not attribute process fluctuations that can be absorbed by the reflow wetting process to the following: When there is insufficient solder coverage, abnormally reduced solder joint height, or failure to form continuous wetting at the solder joint after reflow, establish a causal link between low solder paste volume and the risk of cold solder joints after reflow; when the solder paste volume is high and solder connections occur between adjacent pads after reflow, establish a causal link between excessive solder paste and bridging defects; when the solder paste volume is high but no bridging occurs, and solder balls, local accumulation, or solder overflow are present at the solder joints after reflow, record this target as a low-confidence defect target, and determine whether to include it in the review list based on the aggregated confidence status during confidence correction.

[0090] After verifying the self-alignment convergence relationship, the confidence level of the single-station defect judgment is corrected. When the post-reflow optical inspection alone determines it as a defect, but the cross-process evidence chain shows that the upstream deviation is within the tolerable deviation margin, and the post-reflow position change conforms to the reflow self-alignment convergence relationship, if the aggregate confidence level is high, the final joint judgment is adjusted to non-defect; if the aggregate confidence level is medium or low, it is included in the low-confidence defect review target. When the post-reflow optical inspection alone determines it as qualified, but the cross-process evidence chain shows that there is excessive solder paste, excessive mounting offset, or abnormal overlap in the upstream, and the post-reflow forming state shows signs of deterioration corresponding to the upstream deviation, if the aggregate confidence level is high, the final joint judgment is adjusted to defect; if the aggregate confidence level is medium or low, it is included in the low-confidence qualified review target. When the single-station judgment and the cross-process evidence chain support each other, the target is recorded as a high-confidence judgment target.

[0091] For targets with broken evidence chains, a downgrade judgment is made based on the available sub-chains, and the confidence level is recorded as medium confidence or low confidence. When the broken point is located in key workstations such as furnace front inspection or furnace back inspection and cannot be supplemented by adjacent nodes, the target is included in the manual review list. For targets with collection anomalies, if the furnace back morphology shows tombstone erection, severe offset, or bridging at the same time, the collection anomaly is used as auxiliary evidence of defects. For targets with abnormal image quality, re-acquisition is triggered. If a valid image cannot be obtained after re-acquisition, the target, along with the cross-process evidence chain and anomaly description, is submitted for manual review.

[0092] After the above processing, the following results are generated: confidence-corrected defect judgment results, non-defect judgment results, low-confidence target list, and evidence connection relationships supporting the judgment. These are used as inputs for root cause backtracking and parameter compensation feedback.

[0093] Specifically, such as Figure 6 As shown: Defects confirmed after confidence correction enter the process root cause backtracking and parameter compensation processing; the input data includes at least the final defect judgment result, cross-process evidence chain, deviation direction of each node, deviation amount, tolerable deviation margin, anomaly mark, aggregation confidence status, and process parameter records; during processing, starting from the post-reflow inspection node, backtracking upstream along the process sequence to the reflow process node, pre-reflow inspection node, mounting node, solder paste inspection node, and printing process node, to find the workstation where the first deviation from the tolerable deviation margin occurs and can explain the post-reflow defect type; the workstation with the first deviation is regarded as the main cause process, and the workstations that continue to deviate and have an amplifying effect on the post-reflow defect are regarded as collaborating factors; the workstation with the first deviation is not directly determined by the workstation with the earliest data, but is judged by whether there is a correspondence between the deviation direction and deviation amount of the workstation and the post-reflow defect type;

[0094] For example, if the solder paste volume is less than 50% of the nominal volume at the solder paste inspection node, and insufficient wetting, insufficient solder, or cold solder joints are found at the post-reflow inspection node, the printing process is identified as the main causative process. If the placement node shows a placement point offset exceeding 50% of the pad width, and continuous offset or tombstoning is found at the post-reflow inspection node, the placement process is identified as the main causative process. If both the upstream solder paste state and the placement state are within the tolerable deviation margin, but cold solder joints, poor wetting, or abnormal solder joint appearance occur at the post-reflow stage, and the peak temperature, heating rate, or time above the liquidus line in the reflow process node deviates from the corresponding process window, the reflow process is identified as the main causative process. The aforementioned 50% corresponds to the lower limit of the solder paste volume process window and the tolerance boundary for surface mount component placement, respectively, and is used to distinguish between severe deviations and process fluctuations that can be absorbed by the reflow wetting or self-alignment process.

[0095] For root causes in the printing process, parameter compensation is generated based on the direction and amount of solder paste deviation. When multiple pads in the same area show consecutively low solder paste volumes, the average low-volume ratio is calculated. The average low-volume ratio is the ratio of the difference between the nominal volume and the measured volume to the nominal volume. When the average low-volume ratio exceeds 20%, the area is identified as having a printing under-pour trend. The 20% setting is based on distinguishing between single-point detection fluctuations and regional printing offsets. This value is lower than the severe solder shortage boundary corresponding to the lower limit of the basic process window for solder paste volume, making it suitable as an early compensation trigger condition. When generating compensation information for the printer, the compensation information includes at least the pad area and the deviation ratio. For example, fields such as deviation direction, squeegee pressure adjustment direction, printing speed adjustment direction, demolding speed adjustment direction, and stencil cleaning prompts are included. The compensation range can be generated as a mapping coefficient based on 20% to 50% of the average under-proportion of continuous boards, and converted into the adjustment amount of squeegee pressure, printing speed, or demolding speed according to the allowable range of the printer parameters. This mapping coefficient is used to convert a portion of the actual deviation into a compensation amount to reduce the risk of over-compensation caused by response lag and batch fluctuations in the printing process. When the solder paste volume is too high and the bridging is concentrated on adjacent fine-pitch pads, compensation information for stencil cleaning, squeegee pressure reduction, or printing speed adjustment is generated first.

[0096] For the root cause of the placement process, parameter compensation is generated based on the component placement offset direction, offset magnitude, and rotation angle. When the same component reference number in three consecutive printed circuit boards or ten printed circuit boards in the same batch shows a placement offset in the same direction, it is identified as placement coordinate drift. Among them, three consecutive printed circuit boards serve as a short-cycle window to identify short-term continuous offsets and exclude occasional errors on a single board. Ten printed circuit boards in the same batch serve as a batch window to identify batch repetitive offsets and reflect stability deviations caused by nozzle, visual recognition, or board positioning. When generating compensation information for the placement equipment, the compensation information includes at least the fields of nozzle number, component reference number, average offset vector, average angle deviation, and placement compensation direction. The placement compensation amount can be selected from 50% to 80% of the average offset, and compensation is generated in the opposite direction of the average offset vector. This range is used to retain a safety margin while correcting major deviations, avoiding reverse offsets caused by the coupling of nozzle status, visual recognition errors, and board positioning errors. If the rotation angle deviation changes continuously in the same direction, a visual alignment check or nozzle posture calibration prompt is generated simultaneously.

[0097] For reflow process root causes, parameter compensation is generated based on abnormal solder joint morphology and reflow temperature records after reflow. For tin-silver-copper lead-free solder processes, the peak temperature can be selected from 235℃ to 250℃, and the time above the liquidus line can be selected from 45s to 90s. The above range is based on the reflow process window for commonly used tin-silver-copper lead-free solders to achieve sufficient wetting, metallurgical bonding, and avoid thermal damage to components. The actual range can be calibrated according to solder type, PCB thickness, component heat resistance rating, and product reliability requirements. If the defects after reflow are insufficient wetting, cold soldering, or insufficient solder spreading, and the reflow record shows a low peak temperature or a short time above the liquidus line, then adjustments will be made to increase the peak temperature, extend the time above the liquidus line, or verify the heating rate. If the defects after reflow are solder balls, bridging, or solder overflow, and the amount of solder paste is not significantly exceeded, then adjustments will be made to the preheating time, heating rate, peak temperature curve, or reflow oven temperature uniformity.

[0098] When deviations exist at multiple workstations, for example, when the solder paste volume is close to the upper limit of the tolerable deviation margin, and the placement offset is close to 50% of the pad width, and bridging occurs after the reflow oven, the printing process is identified as the primary cause process, and the placement process is marked as a contributing factor. The compensation information includes reducing the solder paste volume and verifying the placement point. For targets with broken evidence chains, abnormal data collection, or manual review intervention, if the root cause of the defect cannot be uniquely determined, automatic compensation information is not directly issued. Instead, a root cause candidate list and evidence chain summary are output, and parameter adjustments are performed after confirmation by process engineers. The root cause candidate list includes at least the following fields: candidate process, corresponding evidence node, deviation direction, deviation amount, confidence level, and suggested review items.

[0099] After compensation, the occurrence rate of similar defects is continuously tracked, and statistics are performed using the short cycle window or batch window. If the occurrence rate of similar defects decreases, the compensation parameters are retained. If the occurrence rate of similar defects does not decrease, or if abnormalities such as excessive solder paste, reversed landing point, or reflow overheating occur in the opposite direction of the original deviation, the previous stable parameters are restored, and the compensation is recorded as invalid. When the same type of deviation in the same process occurs repeatedly in three consecutive printed circuit boards, or when the proportion of similar deviations in the same shift reaches 5% to 10% of the target number of the process, a process drift warning is generated. Three consecutive printed circuit boards are used to identify short-cycle continuous drift, and 5% to 10% are used to identify shift-level repetitive abnormalities. This proportion is used for process drift warnings and is not used as a criterion for judging whether a single board is qualified or not. The root cause of the defect, the direction of deviation, the amount of deviation, the compensation information, and the compensation verification results are written into the batch-level process record as the basis for subsequent review of similar defects and updating of process parameters.

[0100] Example 2: Based on Example 1, the specific application process of a method for detecting defects before and after SMT assembly is further explained:

[0101] The inspection process of a double-sided mixed-assembly printed circuit board is illustrated as an example. The board-level barcode of this printed circuit board is PCB20260526001. There are three global reference markers on the board surface, and local reference markers are set near the fine-pitch integrated circuit area and the BGA area. The printed circuit board has a chip resistor R105, a small outline integrated circuit U12, and a BGA package device U31. R105 corresponds to two end pads, U12 corresponds to multi-pin pads on both sides, and U31 corresponds to an array of ball pads. The printed circuit board goes through solder paste inspection, mounting, pre-reflow optical inspection, reflow soldering, and post-reflow optical inspection in sequence. The inspection data of each station are classified into the same batch record according to the board-level barcode.

[0102] When the printed circuit board enters the solder paste inspection station, the board-level barcode PCB20260526001 is read, and three global reference markers on the board surface are identified. The solder paste inspection station establishes a conversion relationship between the current image coordinates and the unified reference system based on the measured coordinates and design coordinates of the reference markers. After conversion, the reference point registration residual in the ordinary pad area is 18μm, and the local reference point registration residual in the U12 fine pitch area is 14μm, both of which meet the registration requirements of the corresponding areas. Subsequently, the solder paste inspection data, placement equipment records, pre-reflow optical inspection data, and post-reflow optical inspection data are all recorded according to this unified reference system, including pad position, component position, offset direction, offset amplitude, and acquisition time, so that the data from each station can be attributed to the same printed circuit board and the same physical target.

[0103] For the chip resistor R105, the design data corresponds to the left pad R105-L and the right pad R105-R, with the two end pads forming the same pad cluster. The solder paste inspection station detected that the solder paste volume of pad R105-L was 86% of the nominal volume, the solder paste area was 92% of the nominal area, and the solder paste height was 94% of the nominal height. The solder paste deposition center was offset to the right by 0.04mm relative to the pad center. The solder paste volume of pad R105-R was 91% of the nominal volume, and the solder paste deposition center was offset to the right by 0.04mm relative to the pad center. The solder paste volume was 0.03mm to the right; the volume of the solder paste was within the basic process window of 50% to 150% of the nominal volume, and no solder paste overflow was detected; the mounting equipment record showed that the center of the R105 component landing point was offset to the right by 0.18mm relative to the nominal center of the pad cluster, and the rotation angle deviation was 3.2°; the pre-reflow optical inspection station detected that the two solder ends of R105 still covered the corresponding pads, and the effective overlap ratio between the solder ends and the pads was 56%. The pre-reflow optical inspection results marked the component as a suspected defect of positional offset;

[0104] After reflow soldering, the post-reflow optical inspection station collects the forming state of the solder joints at both ends of R105. The post-reflow inspection results show that the offset of the center of R105 component relative to the nominal center of the pad cluster has been reduced to 0.05mm. The actual displacement direction is from the offset position before reflow to the center of the pad. The wetting boundary of the two end solder joints is continuous, the solder joint spread shape is complete, and no tombstoning, insufficient solder, or bridging is detected. Based on the physical target identification of R105, the solder paste inspection node, placement node, pre-reflow inspection node, and post-reflow inspection node are associated according to the process sequence to obtain the offset change trajectory of the component. The offset change trajectory shows that the pre-reflow offset direction is consistent with the post-reflow convergence direction, the post-reflow offset amplitude is smaller than the pre-reflow offset amplitude, and the forming state of the solder joints after reflow is normal. Therefore, the suspected defect of position offset given by the pre-reflow optical inspection is corrected to a non-defect, and the physical target corresponding to R105 is recorded as a high-confidence non-defect target.

[0105] For the small outline integrated circuit U12, the design data shows that multiple pin pads are set on both sides of U12. The pad cluster identification records the pin arrangement order on the same side, the spacing between adjacent pins, and the correspondence of pins on the opposite side. The solder paste inspection station detected that the solder paste volume of the pad corresponding to the 7th pin of U12 was 166% of the nominal volume, and the solder paste deposition center was offset by 0.05mm towards the adjacent 8th pin. The solder paste volume of the adjacent pads on the same side was 148% of the nominal volume, close to the upper limit of the process window. The mounting equipment records show that the overall landing point offset of U12 was 0.04mm, and the rotation angle deviation was 1.1°. The pre-reflow optical inspection did not detect any obvious pin suspension or severe offset. Therefore, the pre-reflow inspection node did not form the main abnormal evidence that could explain the post-reflow bridging.

[0106] After reflow, the post-reflow optical inspection station detected solder continuity between pins 7 and 8 of U12. The solder joint spread boundary crossed the gap between adjacent pads, forming a bridging defect. After incorporating this target into the cross-process evidence chain, the evidence chain showed that pin 7 of U12 had solder paste volume exceeding the upper limit and the deposition center shifted towards the adjacent pad at the solder paste inspection node. However, no severe shift sufficient to explain the bridging occurred at the mounting node or the pre-reflow inspection node. Tracing back upstream along the process sequence, the first station that could explain the post-reflow bridging defect was the solder paste inspection station, and the root cause was attributed to the printing process. For this root cause, compensation information for the printer was generated. The compensation information included at least the pad area where pin 7 of U12 was located, the proportion of solder paste volume exceeding the upper limit, the direction of solder paste shift, stencil cleaning prompts, the direction of squeegee pressure reduction, and the direction of printing speed adjustment. If the same area showed solder paste volume exceeding the upper limit or adjacent pin bridging trend in three consecutive printed circuit boards, a printing process drift warning was generated, and the compensation information was written into the batch-level process record.

[0107] For BGA packaged device U31, the post-reflow optical inspection station detected strong reflections from local solder joints in the edge area of ​​U31, and some solder joints could not be stably extracted. In this case, instead of using the coordinates of a single solder joint as the sole basis for collection, the station reads the identifiable corner locations, edge row / column directions, and the spacing between solder joints in adjacent complete rows / columns. If at least one corner location or two mutually perpendicular edge row / column directions can be confirmed, the post-reflow solder joint group is assisted in collection to the BGA pad array corresponding to U31, and the collection status is recorded as low-confidence collection. For low-confidence collection targets, defect judgment is not directly used as a basis for high-confidence automatic collection. Instead of focusing on output, the degradation judgment is made by combining solder paste detection nodes, reflow process nodes, and post-reflow forming nodes. If there is no obvious abnormality in solder paste amount at the solder paste detection node and no obvious landing point offset at the placement node in the area corresponding to U31, but the reflow record shows that the time above the liquidus line is lower than the process window, and multiple solder joints are insufficiently wetted after reflow, then the reflow process sequence is listed as the root cause candidate process, and verification information on peak temperature, time above the liquidus line, or temperature uniformity is generated. If the BGA corner point and edge row and column direction cannot be confirmed, it is recorded as an array aggregation anomaly, and the target, along with the post-reflow image and evidence chain summary, is submitted for manual review.

[0108] During this inspection process, different physical targets on the same printed circuit board form independent cross-process evidence chains. The evidence chain corresponding to R105 is used to characterize the convergence process of upstream offset after reflow and correct suspected defects in the pre-reflow single station to non-defects. The evidence chain corresponding to U12 is used to characterize the correspondence between the post-reflow bridge and the solder paste detection node for excess solder paste and deposition offset, and to locate the root cause to the printing process. The evidence chain corresponding to U31 forms root cause candidates in combination with the reflow process record under low confidence aggregation state, and is transferred to manual review when the data is incomplete. Thus, the deviation direction, deviation amount, confidence state, root cause process, and compensation information are written into the batch-level process record as the basis for subsequent review of similar defects and updating of process parameters.

[0109] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0110] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented in software, the above embodiments can be implemented in whole or in part by a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions of the embodiments of this application are implemented in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted wirelessly or wiredly from one website, computer, server, or data center to another website, computer, server, or data center. Wired methods include optical fiber, twisted pair, coaxial cable, etc. Wireless methods include infrared, microwave, etc. Available media include any available media that can be accessed by a computer or data storage devices such as servers and data centers that contain one or more sets of available media. Available media can be magnetic media (floppy disks, hard disks, magnetic tapes), optical media (DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0111] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0112] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for detecting defects before and after SMT assembly, characterized in that, include: S1: Read the printed circuit board level barcode, identify the reference mark points on the board surface, and establish a unified reference system that runs through solder paste inspection, mounting, pre-reflow optical inspection and post-reflow optical inspection; S2: Convert the pad positions, component positions, solder paste status, mounting positions, pre-reflow overlap status and post-reflow forming status collected from each workstation to a unified reference system, and establish physical target identification based on board-level barcode, pad number, component affiliation, package type and relative orientation of pad clusters. S3: Based on the physical target identification, collect the same solder joint or the same component before and after reflow, and attach the status information and deviation information of each station according to the process sequence; S4: Based on upstream deviation information, package type, pad size, solder paste wetting state and post-reflow forming state, verify the reflow self-alignment convergence relationship and correct the single-station defect judgment confidence state. S5: For defects confirmed after confidence correction, trace back upstream along the process sequence to determine the first deviation from the workstation, and feed back the direction and amount of deviation to the corresponding process for parameter compensation.

2. The method for detecting defects before and after SMT assembly according to claim 1, characterized in that, S1 includes: Collect board-level barcodes, board surface images, nominal coordinates of reference markers, and measured coordinates of reference markers. When barcode reading fails, generate a temporary board identifier based on the vehicle number and transit time. Brightness equalization and edge enhancement are performed on the board image. The contour features of three non-collinear global reference markers and local reference markers are identified. Two-dimensional similarity transformation or affine transformation relationship is established based on the same reference markers. Registration residuals are calculated by region, and reference system number, coordinate conversion relationship, residual record, valid status and anomaly mark are generated. When the identification of the reference marker fails, the image is resampled or a backup local reference marker is enabled. When the local residual exceeds the limit, a low confidence spatial state is set.

3. The method for detecting defects before and after SMT assembly according to claim 1, characterized in that, S2 includes: Based on a unified reference system number and coordinate conversion relationship, coordinate transformation, time recording standardization, and board orientation correction are performed on the data of each workstation. Perform mirror calibration on the double-sided panel and correct the station overtime record based on the first code reading time and the production line cycle window. Based on board-level barcode, component tag number, pad number, package type, pad size, nominal center position of pad, and relative distance, relative direction and arrangement order within the pad cluster, a physical target identity record is established, binding the solder paste state, mounting position, pre-reflow overlap state and post-reflow forming state to the corresponding pad or component. When names are inconsistent, tag number mapping is performed; when fields are missing, encapsulation matching, identity confidence marking, and identity anomaly recording are performed.

4. The method for detecting defects before and after SMT assembly according to claim 3, characterized in that, Perform mirror calibration on the double-sided panels, and correct the over-station records based on the first code reading time and production line cycle time window, including: Establish board orientation marks for the front and back sides respectively, and perform mirror conversion of the back coordinates based on the flip-board relationship; The timing anchor point is the first time the solder paste is read at the inspection station. The production line cycle window is determined based on the continuous station pass records, conveyor speed and board spacing. The window width is less than half of the minimum time interval between adjacent boards. The subsequent station pass records are then corrected according to the production line cycle window. Update window parameters when production line speed or board spacing changes.

5. The method for detecting defects before and after SMT assembly according to claim 1, characterized in that, S3 includes: Read the target record of the same board based on the physical target identity table and the workstation data standardization table; Before reflow, the components are associated with the pad number, component affiliation and relative orientation of the pad cluster. After reflow, the overall matching is performed according to the relative distance, relative direction, arrangement order, adjacent spacing ratio and endpoint orientation within the solder joint group. Auxiliary aggregation of BGA arrays is performed based on corner orientation, edge rows and columns, and adjacent complete rows and columns; Establish a chain of evidence based on printing process, solder paste inspection, mounting, pre-reflow inspection, reflow process, and post-reflow inspection, and record the collection confidence status, data breakpoints, barcode mixed reading correction results, and collection anomalies.

6. The method for detecting defects before and after SMT assembly according to claim 5, characterized in that, Auxiliary aggregation of BGA arrays is performed based on corner orientation, edge rows and columns, and adjacent complete rows and columns, including: For BGA solder joint arrays, when some solder joints are not effectively extracted due to obstruction, reflection, insufficient solder, or recognition failure, the orientation of the identified corner points, the direction of the edge rows and columns, and the spacing between adjacent complete rows and columns are read. After confirming the location of a corner point or the row and column directions of two mutually perpendicular edges, array assignment matching is performed. If the corner point orientation and edge row / column direction are not confirmed, or if the relative relationship of solder joint groups does not match the relative relationship of nominal pad clusters, it is recorded as a BGA array aggregation anomaly.

7. The method for detecting defects before and after SMT assembly according to claim 1, characterized in that, S4 includes: Determine the tolerable deviation margins for solder paste quantity, placement offset, rotation angle, and package type based on the cross-process evidence chain; The self-alignment status is verified by the offset direction in front of the furnace, the displacement direction behind the furnace, and the change in offset behind the furnace. By combining the wetting boundary after the furnace, solder joint height, solder continuity, solder balls, tombstoning, and aggregated confidence status, the conclusions of single-station defects are jointly judged, confidence marked, downgraded, and image re-sampling or manually reviewed.

8. The method for detecting defects before and after SMT assembly according to claim 7, characterized in that, The self-alignment status is verified by the offset direction in front of the furnace, the displacement direction behind the furnace, and the change in offset behind the furnace, including: For placement offset targets, read the offset direction and offset magnitude of the placement node or the pre-reflow detection node, and take the direction from the offset position to the center of the pad as the expected convergence direction; Read the final position of the detection node after the furnace, determine the displacement direction and offset amplitude after the furnace, calculate the difference between the offset amplitude before the furnace and the offset amplitude after the furnace, and perform self-alignment status verification based on the angle between the expected convergence direction and the displacement direction after the furnace, the change in offset amplitude, and the abnormal markers of insufficient wetting, bridging, and tombstoning.

9. The method for detecting defects before and after SMT assembly according to claim 1, characterized in that, S5 includes: Based on the final defect determination results, the cross-process evidence chain, and the deviation information of each node, the reflow process node, the pre-reflow process node, the mounting node, the solder paste inspection node, and the printing process node are traced back sequentially from the post-reflow inspection node. The main causative process and cooperating factors are determined based on the correspondence between deviation and post-furnace defect type, and compensation information for printing, mounting or reflow processes is generated respectively. When the root cause is not unique, a list of root cause candidates is output. After manual confirmation, parameter adjustments are performed, the occurrence rate of similar defects is tracked, process drift warnings are generated, and invalid compensation records and parameter rollbacks are recorded.

10. The method for detecting defects before and after SMT assembly according to claim 9, characterized in that, Based on the correspondence between deviations and post-reflow defect types, the main causative processes and cooperating factors are determined, and compensation information for printing, mounting, or reflow processes is generated, including: Corresponding solder paste volume deviation and solder paste deposition offset to insufficient solder, cold solder joint or bridging type, generating printing adjustment information, the printing adjustment information records pad area, deviation ratio, deviation direction, stencil cleaning prompt, squeegee pressure adjustment direction and printing speed adjustment direction. The landing point offset and rotation angle are correlated with the continuous offset or tombstone type to generate placement compensation information. The placement compensation information records the nozzle number, component number, average offset vector and landing point compensation direction. By correlating peak temperature, time above the liquidus line, and heating rate with the types of poor wetting, solder balls, or solder overflow, reflow adjustment information is generated. When multiple workstations deviate, the primary cause process is determined by the workstation with the first deviation, and subsequent workstations with deviations are marked as contributing factors.

Images