A system and method for real-time detection of bond quality

Abstract

The application relates to the technical field of detection, in particular to a bonding quality real-time detection system and method. The system comprises a bonding platform which is independently fixedly arranged; a fixing clamp for fixing a workpiece is arranged on the bonding platform, and a displacement mechanism for driving the fixing clamp to move is arranged on the bonding platform; a bonding arm is arranged above the bonding platform and moves up and down; a bonding wedge is arranged at the tail end of the bonding arm; a line laser scanning unit is fixed above the bonding platform and is used for performing three-dimensional contour scanning on the workpiece and a welding point; a motion control system comprises a servo motor and a grating ruler and is used for driving the bonding arm, the displacement mechanism and the line laser scanning unit; a processor is electrically connected with the bonding arm, the line laser scanning unit, the motion control system and the displacement mechanism, is used for controlling the actions of the components, collecting three-dimensional contour scanning data of the workpiece and the welding point in real time, and calculating a detection result according to the collected data. The application can judge the quality of a bonding point based on multi-dimensional features during the bonding process.

Description

Technical Field

[0001] This invention relates to the field of detection technology, and in particular to a real-time detection system and method for bonding quality. Background Technology

[0002] Wire bonding is a critical process in semiconductor packaging, using metal wires (such as gold or copper wires) to achieve electrical connections between the chip and the substrate pins. The sheer number of bonding points on a single chip places extremely high demands on the consistency of bonding quality. Currently, the industry commonly uses offline sampling inspection for quality monitoring, where a small number of samples are inspected after a batch of products is completed, using an optical microscope or push-pull force tester. This post-production inspection method has the following drawbacks: it cannot detect abnormal fluctuations in the production process in real time, leading to batch defects; the limited number of samples tested makes it difficult to reflect the overall quality distribution; and it lacks quantitative analysis of the three-dimensional morphology of the solder joints, making it difficult to identify potential defects such as cold solder joints, solder joint collapse, and shape distortion based solely on two-dimensional images. Summary of the Invention

[0003] This invention provides a real-time bonding quality detection system and method, which can collect three-dimensional information of the solder joint in real time during the bonding process and make quality judgment based on multi-dimensional features, effectively solving the problems in the background art.

[0004] This invention provides a real-time bonding quality detection system, comprising:

[0005] The bonding platform is independently and fixedly set; the bonding platform is equipped with a fixing fixture for fixing the workpiece, and a displacement mechanism for moving the fixing fixture;

[0006] A bonding arm is located above the bonding platform and moves vertically; a bonding cutter is provided at the end of the bonding arm.

[0007] The line laser scanning unit is fixed above the bonding platform and is used to perform three-dimensional contour scanning of the workpiece and the weld joint.

[0008] The motion control system, including servo motors and grating rulers, is used to drive the bonding arm, displacement mechanism, and line laser scanning unit.

[0009] The processor is electrically connected to the bonding arm, line laser scanning unit, motion control system, and displacement mechanism. It is used to control the actions of each component, acquire three-dimensional contour scanning data of the workpiece and weld points in real time, and calculate the detection results based on the acquired data.

[0010] Furthermore, the line laser scanning unit includes a line laser emitter and a camera, and the laser emission direction of the laser emitter forms an angle with the orientation of the camera.

[0011] Furthermore, multiple heating units are installed at the bottom of the bonding platform to heat the bonding environment.

[0012] This invention also provides a method for real-time detection of bonding quality, comprising the following steps:

[0013] S1: Before bonding, laser scanning is performed on the workpiece area to obtain the reference three-dimensional point cloud data of the workpiece surface;

[0014] S2: After bonding, the workpiece area is laser scanned again to obtain the surface three-dimensional point cloud data including the weld points;

[0015] S3: Perform rigid registration on the two sets of point cloud data before and after the combination, and then calculate the height of each pixel relative to the workpiece by point-by-point difference;

[0016] S4: Based on the height distribution, the solder joint region is extracted by adaptive threshold segmentation, and basic geometric features are calculated from the solder joint region;

[0017] S5: Combining basic geometric features and the original volume of the weld point, construct a theoretical model of the weld point shape, and calculate the volume deviation rate and shape deviation degree between the actual weld point and the theoretical spherical cap.

[0018] S6: Perform surface fitting on the height data of the solder joint area, extract the residual field, and calculate the statistical characteristics of the residual field and the higher-order statistics of the local curvature distribution to evaluate the surface roughness of the solder joint surface.

[0019] S7: Extract the edge contour of the solder joint area, calculate the Fourier descriptor of the edge contour, construct the edge irregularity index based on the energy distribution of the Fourier spectrum, and calculate the overlap rate between the edge contour and the ideal contour.

[0020] S8: Calculate the comprehensive quality index by using a multi-dimensional fusion model to obtain multiple parameters from steps S5 to S7, and determine the quality level of the solder joints according to a preset threshold. Mark and alarm for defective solder joints in real time.

[0021] Furthermore, the height distribution of the weld point relative to the workpiece is calculated using point-by-point difference calculation as follows:

[0022] The reference 3D point cloud of the workpiece surface obtained before bonding is represented as:

[0023] {p pre,i =(x pre,i ,y pre,i ,z pre,i )|i=1,2,……,N};

[0024] Where, p pre,i This represents a 3D vector of the i-th pixel, where the x-coordinate of the i-th pixel in the plane is x. pre,iThe vertical coordinate of the plane is y pre,i The height is z pre,i N is the total number of pixels;

[0025] The 3D point cloud representation of the workpiece surface, including the solder joints, obtained after bonding is as follows:

[0026] {p post,i =(x post,i ,y post,i ,z post,i )|i=1,2,……,N};

[0027] p post,i This represents a 3D vector of the i-th pixel, where the x-coordinate of the i-th pixel in the plane is x. post,i The vertical coordinate of the plane is y post,i The height is z post,i N is the total number of pixels;

[0028] The displacement t in the horizontal direction of the plane is obtained from the displacement of the workpiece before and after bonding. x Displacement t in the vertical direction of the plane y Displacement t in the height direction z ;

[0029] The correspondence between the two sets of 3D point clouds is obtained as follows: x post,i =x pre,i +t x y post,i =y pre,i +t y ;

[0030] For the i-th pixel, obtain the corresponding z-axis. pre,i and z post,i The height H(x) of the solder joint i ,y i )for:

[0031] H(x i ,y i )=z post,i +t z -z pre,i .

[0032] Furthermore, let ΔV be the volume deviation rate of the m-th solder joint region in step S5. m , shape deviation degree △k m ;

[0033] Let Rq be the surface roughness of the m-th solder joint region in step S6. m ;

[0034] Let the edge irregularity index of the m-th solder joint region in step S7 be . The overlap rate between the edge profile and the ideal profile is ΔR. m ;

[0035] In step S8, the comprehensive quality index Q of the m-th solder joint region is calculated. m Specifically:

[0036] Calculate the standardized deviation d for each feature. j =|f j -μ j | / σ j j=1,2,……,5;

[0037] Where f1, f2, f3, f4, and f5 are respectively △V m , △k m 、Rq m , and △R m ;

[0038] μ1, μ2, μ3, μ4, and μ5 are respectively ΔV m , △k m 、Rq m , and △R m The reference mean;

[0039] σ1, σ2, σ3, σ4, and σ5 are respectively ΔV m , △k m 、Rq m , and △R m The reference standard deviation;

[0040] ;

[0041] Where, α j Let be the weights of each feature, and all α j The sum of is 1;

[0042] k1, k2, and k3 are the set contribution parameters;

[0043] β is the set bias constant.

[0044] Furthermore, in step S5, the volume deviation rate ΔV of the m-th solder joint region... m and shape deviation △k m The specific calculation steps are as follows:

[0045] Let V0 be the volume of solder used in each bonding operation;

[0046] In step S4, the area A of the solder joint region of the m-th bonding point is obtained. m ;

[0047] Calculate the actual radius of the m-th solder joint area. ;

[0048] Actual volume ;

[0049] In the formula, Ω m Let m be the coordinate set of the solder joint region of the m-th bonding point;

[0050] △A represents the actual area corresponding to a single pixel;

[0051] H(x q ,y q ) is the coordinate (x q ,y q The height corresponding to );

[0052] Calculate the theoretical height H of the m-th solder joint area. th，m :

[0053] ;

[0054] Volume deviation rate △V m =|V m –V0| / V0;

[0055] Shape deviation degree △k m =|Hmax–H th，m | / H th，m ;

[0056] In the formula, Hmax represents all H(x) values ​​within the region of the m-th solder joint. m ,y m The maximum value in ).

[0057] Furthermore, in step S6, the surface roughness of the m-th solder joint region is Rq. m The specific calculation steps are as follows:

[0058] Constructing a fitted quadratic surface model:

[0059] z(x,y)=λ0+λ1·x+λ2·y+λ3·x 2 +λ4·x·y+λ5·y 2 ;

[0060] Where z(x,y) is the height of the pixel with coordinates (x,y);

[0061] Let the coordinates of the q-th pixel within the solder joint region of the m-th bonding point be (x...). q ,y qSubstitute the point and all pixels within a radius of 3 pixels centered on it into the fitted quadratic surface model to form a system of equations, and solve for the values ​​of λ0, λ1, λ2, λ3, λ4, and λ5 corresponding to q using the least squares method.

[0062] Calculate the final fitted height z of the q-th pixel. q =λ0+λ1·x q +λ2·y q +λ3·x q 2 +λ4·x q ·y q +λ5·y q 2 ;

[0063] Calculate the residual r of the q-th pixel. q =H(x q ,y q )-z q ;

[0064] Calculate the surface roughness Rq of the m-th solder joint region. m :

[0065] ;

[0066] Where, N m The number of pixels in the m-th solder joint region;

[0067] Σr q 2 It is the sum of squares of the residuals corresponding to all pixels in the m-th solder joint region.

[0068] Furthermore, in step S7, the edge irregularity index of the m-th solder joint region is: The overlap rate between the edge profile and the ideal profile is ΔR. m The specific calculation steps are as follows:

[0069] Extract the edge contour of the m-th solder joint region to obtain an ordered sequence of edge contour points:

[0070] {(u p ,v p |p=1,2,……,P};

[0071] Take the average x-coordinate of all pixels in the m-th solder joint region as x m,avg The average y-coordinate of all pixels is y m,avg ;

[0072] Convert each edge contour point into a (x) m,avg ,y m,avgThe polar coordinates of the p-th edge contour point are:

[0073] ;

[0074] θ p =arctan[(v p -y m,avg ) / (u p -x m,avg )];

[0075] With all θ p For the independent variable, all ρ p Construct a function ρ(θ) for the dependent variable, and perform a Fourier expansion on ρ(θ):

[0076] ;

[0077] Where a0 represents all ρ p The average value;

[0078] T is the number of terms in the Fourier expansion;

[0079] a t and b t These are the Fourier coefficients;

[0080] Edge Irregularity Index for:

[0081] ;

[0082] The overlap ratio ΔR between the edge profile and the ideal profile m for:

[0083] .

[0084] Furthermore, in step S8, key parameters are determined from the multiple parameters calculated in steps S5 to S7, and trend curves of the changes of each key parameter over time are plotted. When data drift occurs at multiple consecutive solder joints, an early warning signal is issued.

[0085] The technical solution of this invention can achieve the following technical effects:

[0086] This invention integrates a line laser scanning unit with a bonding device to perform high-precision three-dimensional contour scanning of the same workpiece area before and after bonding, obtaining complete morphological data of the weld joints. Based on multi-step image processing and mathematical modeling, multiple feature parameters with clear physical meaning, such as volume deviation, shape deviation, surface roughness, and edge irregularity, are extracted from the weld joints. These parameters are then integrated into a quality index through a multi-dimensional fusion model, enabling real-time quantitative evaluation of the quality of each weld joint. The hardware system achieves seamless integration of the inspection process and the bonding process, completing full inspection of each weld joint without affecting production efficiency. It can promptly identify and warn of defects and risks, thereby greatly improving bonding quality. Attached Figure Description

[0087] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0088] Figure 1 This is a flowchart illustrating a real-time bonding quality detection method. Detailed Implementation

[0089] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0090] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0091] This invention relates to a real-time bonding quality detection system, comprising:

[0092] The bonding platform is independently and fixedly installed, usually mounted on a ground-mounted equipment frame. The equipment frame is made of high-rigidity cast iron or marble to suppress vibration. The bonding platform is equipped with a fixing fixture for fixing the workpiece, as well as a displacement mechanism to move the fixing fixture.

[0093] The bonding arm is located above the bonding platform and moves up and down; a bonding wedge is provided at the end of the bonding arm to perform wire bonding operations.

[0094] The line laser scanning unit is fixed above the bonding platform and is used to perform three-dimensional contour scanning of the workpiece and the weld joint.

[0095] The motion control system includes servo motors and grating rulers, which are used to drive the bonding arm, displacement mechanism and line laser scanning unit. The grating rulers provide real-time feedback of position information.

[0096] The processor, electrically connected to the bonding arm, line laser scanning unit, motion control system, and displacement mechanism, is used to control the actions of each component, acquire real-time three-dimensional contour scanning data of the workpiece and weld joints, and calculate the detection results based on the acquired data. The processor executes one or more program instructions to perform the steps of the real-time bonding quality detection method described below.

[0097] The line laser scanning unit specifically includes a line laser emitter and a camera. The line laser emitter uses a semiconductor laser with a wavelength of 650nm to 850nm, and its power is adjustable. The emitted beam is shaped into a uniformly wide laser line by a cylindrical mirror. The high-speed camera uses a high-resolution CMOS sensor, with a frame rate of no less than 1000fps, and is equipped with a narrow-band filter to suppress ambient light interference. The laser emitter and the camera's optical axis are fixed to a rigid support at a certain angle (usually 20° to 30°) to meet the requirements of laser triangulation measurement.

[0098] It is preferable to set multiple heating units at the bottom of the bonding platform to heat the bonding environment to meet the temperature requirements of thermo-ultrasonic bonding. By placing the heating units at the bottom of the platform, the heat from the heating units will not be directly transferred to the workpiece, but will be conducted through the bonding platform. When the heat passes through the bonding platform, it will be evenly distributed on the bonding platform first, so that the temperature will be more uniform when it is transferred to the workpiece, avoiding local heating of the workpiece.

[0099] This invention also relates to a method for real-time detection of bonding quality, such as... Figure 1 As shown, the process includes multiple steps from S1 to S8. These steps construct a complete logical chain from 3D data acquisition to multi-dimensional feature extraction and comprehensive quality assessment, enabling real-time quantitative evaluation of the quality of each bonding point. The specific content of each step is as follows:

[0100] S1: Before bonding, laser scanning is performed on the workpiece area to obtain the reference three-dimensional point cloud data of the workpiece surface;

[0101] S2: After bonding, the workpiece area is laser scanned again to obtain the surface three-dimensional point cloud data including the weld points;

[0102] S3: Perform rigid registration on the two sets of point cloud data before and after the assembly to eliminate the displacement caused by the workpiece transfer, and then calculate the height of each pixel relative to the workpiece by point-by-point difference.

[0103] S4: Based on the height distribution, the solder joint region is extracted by adaptive threshold segmentation, and basic geometric features are calculated from the solder joint region;

[0104] S5: Combining basic geometric features and the original volume of the weld point, construct a theoretical model of the weld point shape, and calculate the volume deviation rate and shape deviation degree between the actual weld point and the theoretical spherical cap.

[0105] S6: Perform surface fitting on the height data of the solder joint area, extract the residual field, and calculate the statistical characteristics of the residual field and the higher-order statistics of the local curvature distribution to evaluate the surface roughness of the solder joint surface.

[0106] S7: Extract the edge contour of the solder joint area, calculate the Fourier descriptor of the edge contour, construct the edge irregularity index based on the energy distribution of the Fourier spectrum, and calculate the overlap rate between the edge contour and the ideal contour.

[0107] S8: Calculate the comprehensive quality index by using a multi-dimensional fusion model to obtain multiple parameters from steps S5 to S7, and determine the quality level of the solder joints according to a preset threshold. Mark and alarm for defective solder joints in real time.

[0108] In steps S1 and S2, the specific implementation method of line laser scanning is as follows:

[0109] During scanning, a line laser emitter projects a laser line onto the workpiece surface, and a high-speed camera acquires the distorted image of the laser line at a preset angle (usually 20°~30°). Using the principle of laser triangulation, the three-dimensional coordinates of the corresponding points are calculated based on the offset of the laser line in the image. The scanning process employs a step-by-step method; after acquiring each line, a precision displacement mechanism moves the camera one step size Δs (usually set to 2~5μm) perpendicular to the laser line until the entire workpiece area is covered.

[0110] In step S3, due to the influence of workpiece conveying, the workpiece may be displaced between two scans, but the line laser scanning unit will not move. Therefore, the pixel coordinates before bonding and the pixel coordinates after bonding corresponding to a certain point on the workpiece may not be the same, so they need to be registered.

[0111] In step S3, the displacement of the workpiece before and after bonding is calculated using the iterative nearest point algorithm, and then the height distribution of the weld point relative to the workpiece is calculated using point-by-point difference. Specifically:

[0112] The reference 3D point cloud of the workpiece surface obtained before bonding is represented as:

[0113] {p pre,i =(x pre,i ,y pre,i ,z pre,i)|i=1,2,……,N};

[0114] Where, p pre,i This represents a 3D vector of the i-th pixel, where the x-coordinate of the i-th pixel in the plane is x. pre,i The vertical coordinate of the plane is y pre,i The height is z pre,i N is the total number of pixels.

[0115] The 3D point cloud representation of the workpiece surface, including the solder joints, obtained after bonding is as follows:

[0116] {p post,i =(x post,i ,y post,i ,z post,i )|i=1,2,……,N};

[0117] p post,i This represents a 3D vector of the i-th pixel, where the x-coordinate of the i-th pixel in the plane is x. post,i The vertical coordinate of the plane is y post,i The height is z post,i N is the total number of pixels.

[0118] The displacement t in the horizontal direction of the plane is obtained from the displacement of the workpiece before and after bonding. x Displacement t in the vertical direction of the plane y Displacement t in the height direction z ;

[0119] The correspondence between the two sets of 3D point clouds is obtained as follows: x post,i =x pre,i +t x y post,i =y pre,i +t y ;

[0120] For the i-th pixel, obtain the corresponding z-axis. pre,i and z post,i The height H(x) of the solder joint i ,y i )for:

[0121] H(x i ,y i )=z post,i +t z -z pre,i Its physical meaning is the height difference between the i-th pixel on the workpiece before and after bonding; in places without solder joints, this difference should theoretically be 0; while in places with solder joints, this difference is the height of the solder joint.

[0122] In step S4, the specific steps for extracting the solder joint area and calculating its basic geometric features are as follows:

[0123] First, estimate the standard deviation σ of the background noise. noise This can be obtained statistically from the height value of the solderless area at the edge of the pad. Set an initial threshold Th=k. safe ·σ noise , where k safe Typically, the value is 3 to 5. This will satisfy H(x) i ,y i Points with a value greater than Th are marked as suspected solder joints.

[0124] Connectivity analysis is performed on suspected solder joints. Only when the number of suspected solder joints in a connected region exceeds a set threshold will a suspected solder joint in that region be counted as a solder joint. This connected region is then considered a solder joint region, and each solder joint region corresponds to a bonding point on the workpiece. After identifying all connected regions (solder joint regions) that meet the criteria, suspected solder joints that cannot form sufficiently large connected regions are all removed.

[0125] Obtain the area A of the solder joint region at the m-th bonding point. m =N m ·△A,

[0126] N m Let m be the number of pixels in the m-th solder joint region.

[0127] △A represents the actual area corresponding to a single pixel.

[0128] The diameter D of the gold wire in the bonding process and the feed length L for each bonding operation are known. Therefore, the volume of solder used for each bonding operation can be calculated as V0 = D. 2 L / 4.

[0129] Calculate the actual radius of the m-th solder joint area. .

[0130] Using the principles of calculus, the actual volume of the m-th solder joint region can be calculated. ;

[0131] In the formula, Ω m Let m be the coordinate set of the solder joint region of the m-th bonding point;

[0132] △A represents the actual area corresponding to a single pixel;

[0133] H(x q ,y q ) is the coordinate (x q ,y q The height corresponding to ().

[0134] In step S5, the volume deviation rate ΔV of the m-th solder joint region mand shape deviation △k m The specific calculation steps are as follows:

[0135] In wire bonding, the m-th solder joint area should theoretically be the shape of an ideal spherical cap. In the context of solder joint bonding, the spherical cap has the following parameters:

[0136] The radius of the base of the spherical cap (the circular surface in contact with the pad) is r;

[0137] The height of the spherical cap (the vertical distance from the pad plane to the highest point of the solder joint) is h;

[0138] The radius of curvature of the spherical cap (i.e., the original radius of the solder ball) is R.

[0139] The three are related by the following geometric relationship:

[0140] ;

[0141] The volume V of the spherical crown cap The calculation formula is:

[0142] ;

[0143] Will Substituting these values, the formula for calculating the volume of the spherical crown can be transformed into:

[0144] .

[0145] Now, through step S4, the actual radius r of the solder joint area has been obtained. ; and the actual volume V cap = What we need to find now is the unknown quantity h = theoretical height H. th，m Substituting this into the formula above, we obtain a value for H. th，m A cubic equation in one variable can be theoretically solved by transforming it into its standard form and using the quadratic formula. th，m : .

[0146] Calculate the volume deviation rate ΔV m =|V m –V0| / V0;

[0147] The volume deviation rate reflects the accuracy of the solder amount. An excessively large ΔV indicates that there is too much solder (which may cause a short circuit risk), while an excessively small ΔV indicates that there is too little solder (which may cause insufficient strength).

[0148] Shape deviation degree △k m =|Hmax–H th，m | / H th，m ;

[0149] In the formula, Hmax represents all H(x) values ​​within the region of the m-th solder joint. m ,y m The maximum value in );

[0150] Shape deviation measures the degree of deviation between the actual weld joint profile and an ideal spherical cap with the same volume and projected area, Δk. m An excessively large value indicates that the solder joint may have been over- or under-flattened due to improper ultrasonic parameters.

[0151] In step S6, the surface roughness of the m-th solder joint region is Rq. m The specific calculation steps are as follows:

[0152] To evaluate the microscopic undulations of the solder joint surface, local surface fitting is first performed on each point within the solder joint region, constructing a fitted quadratic surface model for each point's local surface fitting:

[0153] z(x,y)=λ0+λ1·x+λ2·y+λ3·x 2 +λ4·x·y+λ5·y 2 ;

[0154] Where z(x,y) is the height of the pixel with coordinates (x,y).

[0155] Let the coordinates of the q-th pixel within the solder joint region of the m-th bonding point be (x...). q ,y q Substitute the point and all pixels within a radius of 3 pixels centered on it into the fitted quadratic surface model to form a system of equations, and solve for the values ​​of λ0, λ1, λ2, λ3, λ4, and λ5 corresponding to q using the least squares method.

[0156] Calculate the final fitted height z of the q-th pixel. q =λ0+λ1·x q +λ2·y q +λ3·x q 2 +λ4·x q ·y q +λ5·y q 2 .

[0157] Calculate the residual r of the q-th pixel. q =H(x q ,y q )-z q ;

[0158] The residual reflects the micro-fluctuations of a local area. A positive value indicates that the point is above the local trend surface, while a negative value indicates that it is below the trend surface.

[0159] Calculate the surface roughness Rq of the m-th solder joint region. m :

[0160] ;

[0161] Where, N m The number of pixels in the m-th solder joint region;

[0162] Σr q 2 It is the sum of squares of the residuals corresponding to all pixels in the m-th solder joint region;

[0163] The smaller the surface roughness, the smoother the solder joint surface; excessive surface roughness may indicate abnormal vibration or uneven solder flow during the bonding process.

[0164] In step S7, the edge irregularity index of the m-th solder joint region is: The overlap rate between the edge profile and the ideal profile is ΔR. m The specific calculation steps are as follows:

[0165] Using the Canny edge detection or eight-neighbor boundary tracking algorithm, the edge contour of the m-th solder joint region is extracted to obtain an ordered sequence of edge contour points:

[0166] {(u p ,v p )|p=1,2,……,P};where P is the total number of edge contour points.

[0167] Take the average x-coordinate of all pixels in the m-th solder joint region as x m,avg The average y-coordinate of all pixels is y m,avg The final pixel (x) m,avg ,y m,avg It can be regarded as the location of the centroid of the m-th solder joint region.

[0168] Convert each edge contour point into a (x) m,avg ,y m,avg The polar coordinates of the p-th edge contour point are:

[0169] ;

[0170] θ p =arctan[(v p -y m,avg ) / (u p -x m,avg )];

[0171] With all θ p For the independent variable, all ρp Construct a function ρ(θ) for the dependent variable, and perform a Fourier expansion on ρ(θ):

[0172] ;

[0173] Where a0 represents all ρ p The average value;

[0174] T represents the number of Fourier expansion terms. The larger the value, the higher the calculation accuracy, but the longer the calculation process will be. The specific value of the number of Fourier expansion terms is set according to the specific accuracy requirements; usually, 10 to 20 is sufficient to cover the main shape features.

[0175] a t and b t These are the Fourier coefficients, and their specific calculation is based on existing technology and will not be elaborated here.

[0176] Edge Irregularity Index for:

[0177] ;

[0178] In the formula, the fundamental frequency term at t=1 (representing the overall offset) is discarded in the numerator, and the high-frequency harmonics at t≥2 are mainly considered, which correspond to the local distortion of the profile. The larger the edge irregularity index, the more irregular the edge is, and there may be burrs or gaps.

[0179] The overlap ratio ΔR between the edge profile and the ideal profile m for:

[0180] ;

[0181] Overlap rate ΔR m The root mean square relative value of the radial deviation from the actual edge point to the ideal circle, the overlap rate ΔR m The smaller the value, the closer the solder joint is to a circle.

[0182] In step S8, the comprehensive quality index Q of the m-th solder joint region is calculated. m Specifically:

[0183] During the equipment debugging phase, 20-30 bonding points that have been confirmed as qualified through offline testing are continuously collected, and the ΔV is statistically analyzed. m , △k m 、Rq m , and △R m Reference mean and reference standard deviation for each feature;

[0184] Calculate the standardized deviation d for each feature. j =|f j -μj | / σ j j=1,2,……,5;

[0185] Where f1, f2, f3, f4, and f5 are respectively △V m , △k m 、Rq m , and △R m ;

[0186] μ1, μ2, μ3, μ4, and μ5 are respectively ΔV m , △k m 、Rq m , and △R m The reference mean;

[0187] σ1, σ2, σ3, σ4, and σ5 are respectively ΔV m , △k m 、Rq m , and △R m The reference standard deviation;

[0188] ;

[0189] Where, α j Let be the weights of each feature, and all α j The sum of is 1, and its specific value can be set according to process experience. For example, setting it to 0.2 indicates that all characteristics are considered equally important.

[0190] k1, k2, and k3 are the set contribution parameters used to adjust the relative importance of linear and interaction terms, for example, k1=0.6, k2=0.2, and k3=0.2.

[0191] β is a set bias constant, such as β=0;

[0192] k1, k2, k3, and β can also be obtained by logistic regression optimization using historical data.

[0193] Volume deviation △V m With shape deviation △k m The product of these values ​​reflects the overall anomaly in solder distribution; surface roughness Rq m With edge irregularity The product reflects the coupling between local microscopic defects and macroscopic contour distortion. Q m The value range is (0,1), with the closer to 0 indicating better quality and the closer to 1 indicating worse quality.

[0194] Set two thresholds, for example, 0.3 and 0.7. If Q mIf the value is ≤0.3, then the m-th bonding point is judged as excellent; if 0.3 < Q m If Q ≤ 0.7, then the m-th bonding point is considered acceptable; if Q m If the value is greater than 0.7, then the m-th bonding point is considered unqualified.

[0195] To adapt to long-term process drift (such as cutting tool wear and material batch variations), the reference mean μ needs to be dynamically updated. j and reference standard deviation σ j When the pass rate of X consecutive solder joints exceeds a threshold (e.g., ≥90%), the data is updated using the characteristics of these pass solder joints. An exponentially weighted moving average is used to update the mean.

[0196] ;

[0197] Where ω is the learning rate, such as 0.05;

[0198] f j bat This represents the average characteristics of the qualified solder joints in the current batch.

[0199] μ j new μ j The new value;

[0200] μ j old μ j The old values.

[0201] Standard deviation σ j Similar updates are possible, but typically involve slower changes (i.e., a smaller value for ω). Template updates ensure the timeliness of the evaluation benchmark.

[0202] In step S8, key parameters are determined from the multiple parameters calculated in steps S5 to S7, and trend curves of each key parameter over time are plotted to help operators monitor process stability. When multiple solder joints show data drift, i.e., the characteristic values ​​of multiple solder joints exceed the control limits or show a clear monotonic trend, an early warning signal can be issued even if a single solder joint is still within the acceptable range, indicating that process parameters may need to be adjusted.

[0203] The following example, using data obtained from a specific run, will provide a detailed explanation of this method:

[0204] The basic settings for the equipment and solder joints are as follows: line laser scanning step size Δs = 5μm; area corresponding to a single pixel ΔA = Δs 2 =25μm 2The gold wire diameter is dw = 100 μm, the length L of the gold wire supplied for each bonding is 2.6 mm, and the theoretical volume of each solder joint is V0 = π(dw / 2). 2 ·L≈20.4×10 6 μm 3 .

[0205] By integrating the projected area, volume, and maximum height of each solder joint using point cloud methods (i.e., steps S1~S4), measurement data for each solder joint can be obtained. For example, with five solder joints, see the table below:

[0206] Table 1. Measured data of five solder joints

[0207] .

[0208] S5: Calculate the volume deviation rate ΔV and shape deviation Δk between the actual weld joint and the theoretical spherical crown.

[0209] In Table 1, the actual volumes V1~V5 of the five weld joints have been obtained. Based on the actual radius formula... The actual radii r1~r5 of the five weld points can also be obtained. Based on the theoretical height H... th，m Calculation formula:

[0210] ;

[0211] The theoretical height of each solder joint can then be calculated. The calculated data for each solder joint are as follows:

[0212] First solder joint: r1≈300.0μm, H th，1 ≈135.2μm;

[0213] Second solder joint: r2≈300.0μm, H th，2 ≈135.7μm;

[0214] The third solder joint: r3≈300.1μm, H th，3 ≈136.2μm;

[0215] The 4th solder joint: r4≈299.9μm, H th，4 ≈135.4μm;

[0216] 5th solder joint: r5≈310.0μm, H th，5 ≈130.5μm.

[0217] The calculation results of the volume deviation rate ΔV and shape deviation Δk of the 5 weld points are shown in the table below:

[0218] Table 2.5 Volume deviation rate and shape deviation of weld points

[0219] .

[0220] S5~S6: The surface roughness of the 5 solder joints is Rq m The edge irregularity index is The overlap rate between the edge profile and the ideal profile is ΔR. m The calculation results are shown in the table below:

[0221] Table 3. Surface roughness, edge irregularity index, and overlap rate of five weld points.

[0222] .

[0223] S8: Overall Quality Index Q m Calculation:

[0224] The reference mean μ of each characteristic is calculated from historical qualified solder joints. j and reference standard deviation σ j Based on the data of qualified solder joints from the previous production round, the reference mean and reference standard deviation of each characteristic were calculated as follows:

[0225] For the j=1th feature ΔV, μ1=0.0014, σ1=0.0005;

[0226] For the j=2th feature Δk, μ2=0.0037, σ2=0.0015;

[0227] For the j=3rd feature Rq, μ3=0.60, σ3=0.08;

[0228] For the j=4th feature μ4 = 0.060, σ4 = 0.008;

[0229] For the j=5th feature ΔR, μ5=0.030, σ5=0.008.

[0230] According to the standardized deviation formula d j =|f j -μ j | / σ j The standardized deviation of each feature was calculated, and the results are shown in the table below:

[0231] Table 4. Standardized Deviation of 5 Features

[0232] .

[0233] Set feature weights α j =[0.30,0.30,0.15,0.15,0.10], contribution parameters k1=0.6, k2=0.2, k3=0.2, bias β=0.

[0234] Based on the interaction term calculation formula and the comprehensive quality index Q m Calculation formula:

[0235] ξ=△V m ·△k m η=Rq m · ;

[0236] ;

[0237] The overall quality index Q of the five solder joints can be calculated. m The results are shown in the table below:

[0238] Table 5. Overall Quality Index of 5 Solder Joints

[0239] .

[0240] Q m The value range is (0, 1), where closer to 0 indicates better quality and closer to 1 indicates worse quality. A threshold Q is set. 不良 With a value of 0.7, solder joints numbered 1 through 4 can be identified as qualified, while solder joint number 5 is deemed unqualified. Reviewing the calculation parameters for solder joint number 5, the unqualified solder joint 5, due to its significantly larger size, rough surface, and irregular edges, had a comprehensive quality index close to 1, thus accurately being judged as defective.

[0241] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of the application as defined herein, and are to be considered as covering any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Thus, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for real-time detection of bonding quality, characterized in that the steps include: include: S1: Before bonding, laser scanning is performed on the workpiece area to obtain the reference three-dimensional point cloud data of the workpiece surface; S2: After bonding, the workpiece area is laser scanned again to obtain the surface three-dimensional point cloud data including the weld points; S3: Perform rigid registration on the two sets of point cloud data before and after the combination, and then calculate the height of each pixel relative to the workpiece by point-by-point difference; S4: Based on the height distribution, the solder joint region is extracted by adaptive threshold segmentation, and basic geometric features are calculated from the solder joint region; S5: Combining basic geometric features and the original volume of the weld point, construct a theoretical model of the weld point shape, and calculate the volume deviation rate and shape deviation degree between the actual weld point and the theoretical spherical cap. S6: Perform surface fitting on the height data of the solder joint area, extract the residual field, and calculate the statistical characteristics of the residual field and the higher-order statistics of the local curvature distribution to evaluate the surface roughness of the solder joint surface. S7: Extract the edge contour of the solder joint area, calculate the Fourier descriptor of the edge contour, construct the edge irregularity index based on the energy distribution of the Fourier spectrum, and calculate the overlap rate between the edge contour and the ideal contour. S8: Calculate the comprehensive quality index by using the multidimensional fusion model to calculate the multiple parameters obtained in steps S5 to S7, and determine the quality level of the solder joint according to the preset threshold. Mark and alarm the defective solder joints in real time.

2. The real-time bonding quality detection method according to claim 1, characterized in that, In step S3, the height distribution of the weld point relative to the workpiece is calculated using point-by-point difference as follows: The reference 3D point cloud of the workpiece surface obtained before bonding is represented as: {p pre,i =(x pre,i ,y pre,i ,z pre,i )|i=1,2,......,N}; Where, p pre,i This represents a 3D vector of the i-th pixel, where the x-coordinate of the i-th pixel in the plane is x. pre,i The vertical coordinate of the plane is y pre,i The height is z pre,i N is the total number of pixels; The 3D point cloud representation of the workpiece surface, including the solder joints, obtained after bonding is as follows: {p post,i =(x post,i ,y post,i ,z post,i )|i=1,2,......,N}; p post,i This represents a 3D vector of the i-th pixel, where the x-coordinate of the i-th pixel in the plane is x. post,i The vertical coordinate of the plane is y post,i The height is z post,i N is the total number of pixels; The displacement t in the horizontal direction of the plane is obtained from the displacement of the workpiece before and after bonding. x Displacement t in the vertical direction of the plane y Displacement t in the height direction z ; The correspondence between the two sets of 3D point clouds is obtained as follows: x post,i =x pre,i +t x y post,i =y pre,i +t y ; For the i-th pixel, obtain the corresponding z-axis. pre,i and z post,i The height H(x) of the solder joint i ,y i )for: H(x i ,y i )=z post,i +t z -z pre,i 。 3. The real-time bonding quality detection method according to claim 1, characterized in that, Let the volume deviation rate of the m-th weld point region in step S5 be ΔV. m , shape deviation degree △k m ; Let Rq be the surface roughness of the m-th solder joint region in step S6. m ; Let the edge irregularity index of the m-th solder joint region in step S7 be . The overlap rate between the edge profile and the ideal profile is ΔR. m ; In step S8, the comprehensive quality index Q of the m-th solder joint region is calculated. m Specifically: Calculate the standardized deviation d for each feature. j =|f j -μ j | / σ j j=1,2,......,5; Where f1, f2, f3, f4, and f5 are respectively △V m , △k m 、Rq m , and △R m ; μ1, μ2, μ3, μ4, and μ5 are respectively ΔV m , △k m 、Rq m , and △R m The reference mean; σ1, σ2, σ3, σ4, and σ5 are respectively ΔV m , △k m 、Rq m , and △R m The reference standard deviation; ； Where, α j Let be the weights of each feature, and all α j The sum of is 1; k1, k2, and k3 are the set contribution parameters; β is the set bias constant.

4. The real-time bonding quality detection method according to claim 3, characterized in that, In step S5, the volume deviation rate ΔV of the m-th solder joint region m and shape deviation △k m The specific calculation steps are as follows: Let V0 be the volume of solder used in each bonding operation; In step S4, the area A of the solder joint region of the m-th bonding point is obtained. m ; Calculate the actual radius of the m-th solder joint area. ; Actual volume ; In the formula, Ω m Let m be the coordinate set of the solder joint region of the m-th bonding point; △A represents the actual area corresponding to a single pixel; H(x q ,y q ) is the coordinate (x q ,y q The height corresponding to ); Calculate the theoretical height H of the m-th solder joint area. th，m : ； Volume deviation rate △V m =|V m -V0| / V0; Shape deviation △k m =|Hmax-H th，m | / H th，m ; In the formula, Hmax represents all H(x) values ​​within the region of the m-th solder joint. m ,y m The maximum value in ).

5. The real-time bonding quality detection method according to claim 3, characterized in that, In step S6, the surface roughness of the m-th solder joint region is Rq. m The specific calculation steps are as follows: Constructing a fitted quadratic surface model: z(x,y)=λ0+λ1·x+λ2·y+λ3·x 2 +λ4·x·y+λ5·y 2 ; Where z(x,y) is the height of the pixel with coordinates (x,y); Let the coordinates of the q-th pixel within the solder joint region of the m-th bonding point be (x...). q ,y q Substitute the point and all pixels within a radius of 3 pixels centered on it into the fitted quadratic surface model to form a system of equations, and solve for the values ​​of λ0, λ1, λ2, λ3, λ4, and λ5 corresponding to q using the least squares method. Calculate the final fitted height z of the q-th pixel. q =λ0+λ1·x q +λ2·y q +λ3·x q 2 +λ4·x q ·y q +λ5·y q 2 ; Calculate the residual r of the q-th pixel. q =H(x q ,y q )-z q ; Calculate the surface roughness Rq of the m-th solder joint region. m : ； Where, N m The number of pixels in the m-th solder joint region; Σr q 2 It is the sum of squares of the residuals corresponding to all pixels in the m-th solder joint region.

6. The real-time bonding quality detection method according to claim 5, characterized in that, In step S7, the edge irregularity index of the m-th solder joint region is: The overlap rate between the edge profile and the ideal profile is ΔR. m The specific calculation steps are as follows: Extract the edge contour of the m-th solder joint region to obtain an ordered sequence of edge contour points: {(u p ,v p )|p=1,2,......,P}; Take the average x-coordinate of all pixels in the m-th solder joint region as x m,avg The average y-coordinate of all pixels is y m,avg ; Convert each edge contour point into a value with (x) m,avg ,y m,avg The polar coordinates of the p-th edge contour point are: ； θ p =arctan[(v p -y m,avg ) / (u p -x m,avg )]; With all θ p For the independent variable, all ρ p Construct a function ρ(θ) for the dependent variable, and perform a Fourier expansion on ρ(θ): ； Where a0 represents all ρ p The average value; T is the number of terms in the Fourier expansion; a t and b t These are the Fourier coefficients; Edge Irregularity Index for: ； The overlap ratio ΔR between the edge profile and the ideal profile m for: 。 7. The real-time bonding quality detection method according to claim 1, characterized in that, In step S8, key parameters are determined from the multiple parameters calculated in steps S5 to S7, and trend curves of each key parameter over time are plotted. When multiple solder joints show data drift, an early warning signal is issued.

8. A real-time bond quality detection system, used to implement the real-time bond quality detection method as described in any one of claims 1 to 7, characterized in that, include: A bonding platform is independently and fixedly set; the bonding platform is equipped with a fixing fixture for fixing the workpiece, and a displacement mechanism for moving the fixing fixture; A bonding arm is located above the bonding platform and moves vertically; a bonding cutter is provided at the end of the bonding arm. A line laser scanning unit is fixed above the bonding platform and is used to perform three-dimensional contour scanning of the workpiece and the weld points. The motion control system includes a servo motor and a grating ruler for driving the bonding arm, the displacement mechanism, and the line laser scanning unit; The processor is electrically connected to the bonding arm, the line laser scanning unit, the motion control system, and the displacement mechanism. It is used to control the actions of each component, acquire three-dimensional contour scanning data of the workpiece and weld points in real time, and calculate the detection results based on the acquired data.

9. The real-time bonding quality detection system according to claim 8, characterized in that, The line laser scanning unit includes a line laser emitter and a camera, and the laser emission direction of the laser emitter forms an angle with the orientation of the camera.

10. The real-time bonding quality detection system according to claim 8, characterized in that, The bonding platform is equipped with multiple heating units at its bottom for heating the bonding environment.

Images