Method and system for adjusting filling process of solid crystal glue

Abstract

The invention provides a filling process adjusting method and system for solid crystal glue, and relates to the technical field of data processing.The method comprises the steps that N filling process nodes are obtained, quality detection results are extracted, then defect authentication is carried out, a sample defect quality detection result cluster is generated, defect degree analysis is carried out, and K defect degrees are obtained, and calling a defect feature-node adaptation network layer, performing node identification on K defect features, respectively matching K parameter adjustment compensations, performing fine adjustment on parameter values, and finally performing process parameter adjustment. According to the invention, the technical problem that the production efficiency and the product yield are further affected due to unstable product quality and easy defects caused by unreasonable process parameter setting and lack of a dynamic adjustment mechanism of the die bond glue filling production line in the prior art can be solved, the quality control and process parameter adjustment flow of the die bond glue filling production line is optimized, and the production efficiency and the product yield are improved. And the product quality and the production efficiency are comprehensively improved.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and particularly to a method and system for adjusting the filling process of die attach glue. Background Art

[0002] In the production of semiconductor devices, die attach glue plays a crucial role. It can bond the chip to a specified area on the bracket or substrate, forming a thermal path or an electrical path, providing necessary conditions for subsequent wire bonding. With the continuous development of technology, higher requirements are put forward for the filling process of die attach glue. In order to ensure good connection between the chip and the substrate and improve the reliability and performance of the product, it is necessary to precisely adjust and control the filling process of die attach glue. With the rapid development of automation and intelligent manufacturing technologies, the filling process of die attach glue is gradually developing towards automation, high precision, and intelligence. Therefore, researching the method for adjusting the filling process of die attach glue is of great significance for improving the production efficiency, reducing costs, and enhancing product quality of semiconductor devices.

[0003] Currently, the process parameters in the prior art are set unreasonably and cannot be adjusted in time. For example, the filling speed, temperature, pressure, etc. are unreasonable, resulting in defects such as uneven filling, bubbles, and voids. These defects will not only reduce the performance and service life of the product but also increase production costs and the scrap rate. At the same time, due to the lack of an effective dynamic adjustment mechanism, the production line cannot adjust the process parameters in a timely manner according to the actual situation, making the problem persist and possibly worsen.

[0004] In summary, the process parameters of the die attach glue filling production line in the prior art are set unreasonably, and there is a lack of a dynamic adjustment mechanism, resulting in unstable product quality, easy occurrence of defects, and further affecting production efficiency and product yield. Summary of the Invention

[0005] The purpose of this application is to provide a method and system for adjusting the filling process of die attach glue to solve the technical problems in the prior art that the process parameters of the die attach glue filling production line are set unreasonably, and there is a lack of a dynamic adjustment mechanism, resulting in unstable product quality, easy occurrence of defects, and further affecting production efficiency and product yield.

[0006] In view of the above problems, this application provides a method and system for adjusting the filling process of die attach glue.

[0007] In a first aspect, the present application provides a method for adjusting the filling process of die bonding glue, which is implemented by a system for adjusting the filling process of die bonding glue. Among them, the method includes: interacting with the die bonding glue filling production line to extract nodes, obtaining N filling process nodes, where the N filling process nodes have N node parameter tolerance bandwidths; extracting the quality inspection results of multiple die bonding glue filling samples within a preset feedback window, and performing defect certification on the extracted quality inspection results of multiple samples according to defect characteristics to generate a sample defect quality inspection result cluster, where the quality inspection results of the multiple samples have multiple defect characteristic sets; taking the defect characteristic type as an index, performing defect degree analysis on the multiple defect characteristic sets in the sample defect quality inspection result cluster and the calibrated defect characteristic value set to obtain K defect degrees of K defect characteristics; calling the defect characteristic-node adaptation network layer to perform node recognition on the K defect characteristics to obtain K associated node sets; respectively matching K parameter adjustment step sizes according to the K defect degrees, and based on the K parameter adjustment step sizes, finely adjusting the current parameter values of the process nodes in the K associated node sets with the N node parameter tolerance bandwidths as constraints to obtain K associated node fine-tuning parameter sets; adjusting the process parameters of the die bonding glue filling production line according to the K associated node fine-tuning parameter sets.

[0008] Second aspect, the present application also provides a filling process adjustment system for die bonding glue, which is used to execute a filling process adjustment method for die bonding glue as described in the first aspect. Wherein, the system includes: a node extraction module, which is used to interact with the die bonding glue filling production line to extract nodes, and obtain N filling process nodes, where the N filling process nodes have N node parameter tolerance bandwidths; a defect certification module, which is used to extract the quality inspection results of multiple die bonding glue filling samples within a preset feedback window, perform defect certification on the extracted multiple sample quality inspection results according to defect characteristics, and generate a sample defect quality inspection result cluster, where the multiple sample quality inspection results have multiple defect feature sets; a defect degree acquisition module, which is used to use the defect feature type as an index to perform defect degree analysis on the multiple defect feature sets in the sample defect quality inspection result cluster and the calibrated defect feature value set, and obtain K defect degrees of K defect features; a node identification module, which is used to call a defect feature-node adaptation network layer to perform node identification on the K defect features, and obtain K associated node sets; a parameter set acquisition module, which is used to respectively match K parameter adjustment step sizes according to the K defect degrees, and based on the N node parameter tolerance bandwidths, fine-tune the current parameter values of the process nodes in the K associated node sets according to the K parameter adjustment step sizes, and obtain K associated node fine-tuning parameter sets; a parameter adjustment module, which is used to adjust the process parameters of the die bonding glue filling production line according to the K associated node fine-tuning parameter sets.

[0009] One or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0010] Node extraction is performed through an interactive die bonding glue filling production line to obtain N filling process nodes. Among them, the N filling process nodes have N node parameter tolerance bandwidths; the quality inspection results of multiple die bonding glue filling samples within a preset feedback window are extracted, and the extracted quality inspection results of multiple samples are defect-certified according to defect characteristics to generate a sample defect quality inspection result cluster. Among them, the quality inspection results of the multiple samples have multiple defect feature sets; using the defect feature type as an index, defect degree analysis is performed on the multiple defect feature sets in the sample defect quality inspection result cluster and the calibrated defect feature value set to obtain K defect degrees of K defect features; the defect feature-node adaptation network layer is called to perform node recognition on the K defect features to obtain K associated node sets; K parameter adjustment step sizes are respectively matched according to the K defect degrees, and based on the N node parameter tolerance bandwidths as a constraint, the current parameter values of the process nodes in the K associated node sets are finely adjusted according to the K parameter adjustment step sizes to obtain K associated node fine-tuning parameter sets; the process parameters of the die bonding glue filling production line are adjusted according to the K associated node fine-tuning parameter sets, effectively solving the technical problems in the prior art that the process parameter settings of the die bonding glue filling production line are unreasonable and lack a dynamic adjustment mechanism, resulting in unstable product quality, easy occurrence of defects, and further affecting production efficiency and product yield. The quality control and process parameter adjustment process of the die bonding glue filling production line are optimized, and the product quality and production efficiency are comprehensively improved.

[0011] The above description is only an overview of the technical solution of the present application. In order to be able to understand the technical means of the present application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of the present application more obvious and understandable, the specific embodiments of the present application are specifically exemplified below. It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present application, nor is it used to limit the scope of the present application. Other features of the present application will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings described below are only exemplary, and for those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0013] Figure 1 It is a schematic flow chart of a method for adjusting the filling process of a die bonding glue of the present application;

[0014] Figure 2 It is a schematic structural diagram of a system for adjusting the filling process of a die bonding glue of the present application.

[0015] Description of reference numerals:

[0016] Node extraction module 11, defect authentication module 12, defect degree acquisition module 13, node identification module 14, parameter set acquisition module 15, parameter adjustment module 16. DETAILED DESCRIPTION

[0017] The present application provides a method and system for adjusting the filling process of a bonding adhesive, thereby solving the technical problems that the process parameter settings of the bonding adhesive filling production line in the prior art are unreasonable and lack a dynamic adjustment mechanism, resulting in unstable product quality, prone to defects, and further affecting production efficiency and product yield. The application optimizes the quality control and process parameter adjustment flow of the bonding adhesive filling production line, thereby comprehensively improving product quality and production efficiency.

[0018] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.

[0019] Embodiment 1

[0020] Please refer to the attached Figure 1 The present application provides a method for adjusting a filling process of a bonding adhesive, wherein the method is applied to a filling process adjustment system of a bonding adhesive, and is also applied to a filling process adjustment system, wherein the filling process adjustment system is communicatively connected to a bonding adhesive filling production line, and the method specifically comprises the following steps:

[0021] S1: The interactive die-bonding adhesive filling production line performs node extraction to obtain N filling process nodes, wherein the N filling process nodes have N node parameter tolerance bandwidths;

[0022] Specifically, the entire filling process is decomposed into N independent process nodes. Each node represents a specific operation link, such as colloid modulation, mold filling, exhaust treatment, etc. For each extracted process node, analyze its key process parameters. These parameters include temperature, time, pressure, flow, etc. Determine the tolerance bandwidth of each parameter. The tolerance bandwidth refers to the range within which the product quality can still be maintained at an acceptable level within this parameter range.

[0023] S2: Extract the quality inspection results of multiple die bonding glue filling samples within a preset feedback window, conduct defect certification on the extracted quality inspection results of multiple samples according to defect characteristics, and generate a sample defect quality inspection result cluster. Among them, the quality inspection results of the multiple samples have multiple defect characteristic sets;

[0024] Specifically, collect the quality inspection results of multiple die bonding glue filling samples from within the preset feedback window. These results include various test data, such as conductive performance tests, appearance inspections, dimensional measurements, etc. Conduct a detailed analysis of the quality inspection results of each sample to identify the defect characteristics therein. Defect characteristics include bubbles, cracks, uneven filling, poor conductive performance, etc. Conduct defect certification on each sample according to the identified defect characteristics. Compare the identified defect characteristics with a known defect characteristic database. The defect characteristic database is a collection containing various known defect types and their typical characteristics, and these characteristics include appearance defects, performance anomalies, etc. According to the comparison results, identify the specific defect types, such as bubbles, cracks, incomplete filling, etc. Classify the identified defects to distinguish major defects and minor defects. Classify the samples with the same or similar defect characteristics into the same quality inspection result cluster. Conduct a set analysis of the defect characteristics in each quality inspection result cluster. Through statistics and analysis, identify major defects, minor defects, and their correlations.

[0025] S3: Using the defect characteristic type as an index, conduct defect degree analysis on the multiple defect characteristic sets in the sample defect quality inspection result cluster and the calibrated defect characteristic value set to obtain the K defect degrees of K defect characteristics;

[0026] Specifically, the defect characteristic value set contains various possible defect types and their corresponding standard defect degree values. These calibrated values can be obtained through historical data and industry standard evaluations and are used as a comparison benchmark. Each defect characteristic set details information such as the type, location, and size of a certain defect. Traverse each defect characteristic set in the sample defect quality inspection result cluster. For each defect characteristic set, using its defect characteristic type as an index, find the corresponding calibrated value from the calibrated defect characteristic value set. Compare the actually detected defect characteristics with the calibrated defect characteristic values. According to the comparison results, conduct defect degree analysis on each defect characteristic. This is achieved by calculating similarity, difference, or other relevant indicators between the actual defect and the calibrated defect. For each type of defect characteristic, assuming there are a total of K types, a specific defect degree value, that is, the defect degree, will be obtained. This defect degree reflects the deviation degree between the actual defect and the calibrated defect and can be used to quantify the severity of the defect.

[0027] S4: Call the defect feature-node adaptation network layer to perform node recognition on the K defect features, and obtain K associated node sets;

[0028] Specifically, call the defect feature-node adaptation network layer, which is a pre-trained machine learning model, such as a deep learning network, that has been trained to identify the relationship between defect features and production line nodes. Input the K defect features and their corresponding defect degrees as input data into the defect feature-node adaptation network layer. Through the calculation and inference of the network layer, identify the production nodes most relevant to each defect feature. The network layer outputs K node sets associated with the K defect features. Each defect feature will correspond to one or more possible production nodes that may cause the defect during the production process.

[0029] S5: Respectively match K parameter adjustment steps according to the K defect degrees, and with the N-node parameter tolerance bandwidth as the constraint, fine-tune the current parameter values of the process nodes in the K associated node sets based on the K parameter adjustment steps to obtain K associated node fine-tuning parameter sets;

[0030] Specifically, for each defect feature, determine the step size of parameter adjustment according to its defect degree. Features with higher defect degrees require larger adjustment steps to significantly improve the process parameters of the relevant nodes. For each process node, there is a parameter tolerance bandwidth, that is, the maximum range within which the node parameters are allowed to be adjusted. This bandwidth is set based on process requirements, equipment capabilities, and safety considerations. Ensure that the parameter fine-tuning does not exceed this tolerance bandwidth to maintain the stability and safety of the production line. For each associated node, fine-tune the current parameter value according to its corresponding defect feature and defect degree, as well as the calculated parameter adjustment step. Record the fine-tuned parameter values of each associated node to form a fine-tuning parameter set of K associated nodes. This set will be used as the new parameter settings for subsequent production or experiments.

[0031] S6: Adjust the process parameters of the die attach glue filling production line according to the K associated node fine-tuning parameter sets.

[0032] Specifically, before any parameter adjustment, it is necessary to back up the current process parameter settings of the production line. Ensure that in case of any problems during the adjustment process, it can be quickly restored to the original state to ensure the stable operation of the production line. Fine-tune the parameter set according to K associated nodes and formulate an adjustment plan. This includes each node that needs to be adjusted, the name of the parameter to be adjusted, the parameter values before and after the adjustment, the adjustment schedule, etc. According to the adjustment plan, gradually adjust the relevant process nodes on the production line. Adjust only one or a few parameters at a time to accurately observe the impact of each adjustment on product quality and production efficiency. During the adjustment process, closely monitor the operating status of the production line and product quality. Record the production data after each adjustment, including key indicators such as product qualification rate, production efficiency, and equipment failure rate. After each adjustment, analyze the collected production data to evaluate the effect of the parameter adjustment. If there is a significant improvement in product quality or production efficiency, it indicates that the adjustment is effective. If the initial adjustment does not achieve the expected effect or new problems occur, iterative optimization needs to be carried out according to the actual situation. This includes further fine-tuning parameters, adjusting strategies, or introducing new improvement measures. When the parameter adjustment achieves the expected effect and after a period of stable operation, the standard operation procedure of the production line should be updated in a timely manner to incorporate the new parameter settings to ensure the stability and consistency of subsequent production.

[0033] Further, step S2 of this application further includes:

[0034] Construct a defect certification space based on a preset defect feature set, where the preset defect feature set includes bubble defects, filling uniformity, and edge integrity;

[0035] Construct multiple spatial coordinate points with the defect feature types and defect values of the multiple defect feature sets, and mark the multiple spatial coordinate points using the quality inspection results of multiple samples.

[0036] Based on the multiple spatial coordinate points and the defect certification space, perform defect certification on the multiple die bonding glue filling samples to obtain a cluster of sample defect quality inspection results.

[0037] Specifically, a preset set of defect features is defined, which includes bubble defects, filling uniformity, color uniformity, presence or absence of foreign objects, size consistency, and edge integrity. A spatial dimension is defined, and in this three-dimensional space, each dimension represents a defect feature respectively. For example, the X-axis can represent the severity of bubble defects, the Y-axis represents filling uniformity, and the Z-axis represents edge integrity. For each die bonding glue filling sample, data regarding bubble defects, filling uniformity, and edge integrity are collected. These data will be converted into spatial coordinate points. For example, the numerical values of the severity of bubble defects, filling uniformity, and edge integrity of a sample can be converted into a point in a three-dimensional space. The spatial coordinate points are labeled using the quality inspection results of multiple samples. The labels may include qualified, unqualified, or other more specific classifications. The spatial coordinate points of each sample are located in the defect certification space. Based on the position and label in the space, some rules are defined for defect certification. For example, if a certain dimension of a point, such as bubble defects, exceeds a certain threshold, then the sample may be considered unqualified. By applying these rules, defect certification can be performed on multiple die bonding glue filling samples, and a cluster of sample defect quality inspection results is obtained. This result cluster will contain the certification status and specific defect information of each sample.

[0038] Further, this application also includes:

[0039] Map multiple spatial coordinate points in the defect certification space to a first-dimensional plane to obtain a first-dimensional scatter plot, where the first-dimensional plane is a plane formed by the first and second coordinate axes of the defect certification space;

[0040] Extract the upper and lower boundaries of the first-dimensional scatter plot, and use the straight line passing through the center of the upper and lower boundaries and parallel to the first coordinate axis as the first-dimensional horizontal starting line;

[0041] Extract the left and right boundaries of the first-dimensional scatter plot, and use the straight line passing through the center of the upper and lower boundaries and perpendicular to the first coordinate axis as the first-dimensional vertical starting line;

[0042] Based on the first-dimensional horizontal starting line and the first-dimensional vertical starting line, perform horizontal distribution recognition and vertical distribution recognition respectively to obtain a first-dimensional defect plane;

[0043] Map multiple spatial coordinate points in the defect certification space to a second-dimensional plane to obtain a second-dimensional scatter plot, and perform defect certification analysis on the second-dimensional scatter plot to obtain a second-dimensional defect plane, where the second-dimensional plane is a plane formed by the second and third coordinate axes of the defect certification space;

[0044] Map multiple spatial coordinate points in the defect certification space to a third-dimensional plane to obtain a third-dimensional scatter plot, and perform defect certification analysis on the third-dimensional scatter plot to obtain a third-dimensional defect plane, where the third-dimensional plane is the plane formed by the first-dimensional coordinate axis and the third-dimensional coordinate axis of the defect certification space;

[0045] Construct a target defect certification subspace based on the first-dimensional defect plane, the second-dimensional defect plane, and the third-dimensional defect plane, and use the multiple sample quality inspection results corresponding to the multiple spatial coordinate points in the target defect certification subspace as a sample defect quality inspection result cluster.

[0046] Specifically, select the first - dimensional coordinate axis of the defect certification space. For example, the plane formed by the severity of bubble defects and the second - dimensional coordinate axis, such as filling uniformity, is used as the first - dimensional plane. Map each coordinate point in the three - dimensional space onto this two - dimensional plane to form a two - dimensional scatter plot, which is called the first - dimensional scatter plot. Extract the upper and lower boundaries and the left and right boundaries from the first - dimensional scatter plot. These boundaries represent the extreme values of the data points in the selected two dimensions. Draw a straight line parallel to the first - dimensional coordinate axis through the center of the upper and lower boundaries as the first - dimensional horizontal starting line. Draw a straight line perpendicular to the first - dimensional coordinate axis through the center of the upper and lower boundaries as the first - dimensional vertical starting line. Based on the first - dimensional horizontal starting line, analyze the distribution of data points in the horizontal direction, that is, the direction of the first - dimensional coordinate axis. Based on the first - dimensional vertical starting line, analyze the distribution of data points in the vertical direction, that is, the direction of the second - dimensional coordinate axis. Through the identification of the horizontal and vertical distributions, the overall distribution of data points on the first - dimensional plane can be obtained, and then possible defect patterns and trends can be identified. This analysis result can be called the first - dimensional defect plane. Select the second - dimensional coordinate axis, such as filling uniformity, and the third - dimensional coordinate axis, such as edge integrity, to form a plane as the second - dimensional plane. Map the coordinate points in the three - dimensional space onto the second - dimensional plane to form a second - dimensional scatter plot. Analyze the second - dimensional scatter plot to identify defect patterns and trends on this plane. According to the analysis result, obtain the defect distribution on the second - dimensional plane, that is, the second - dimensional defect plane. Select the first - dimensional coordinate axis and the third - dimensional coordinate axis of the defect certification space, such as the plane formed by size consistency, as the third - dimensional plane. Map the coordinate points in the three - dimensional space onto the third - dimensional plane to generate a third - dimensional scatter plot. Conduct a detailed analysis of the third - dimensional scatter plot to observe the distribution and trend of data points. Through the distribution of the scatter plot, identify the defect patterns and rules presented on this specific plane. Based on the analysis result, obtain the defect distribution on the third - dimensional plane to form the third - dimensional defect plane. Integrate the analysis results of the first - dimensional defect plane, the second - dimensional defect plane, and the third - dimensional defect plane. According to the analysis results of these three defect planes, a subspace containing the main defect patterns in the three - dimensional space can be determined, that is, the target defect certification subspace. Within this subspace, extract the quality inspection results of multiple samples corresponding to multiple spatial coordinate points. Organize and classify the extracted sample quality inspection results. Finally, obtain a set containing the quality inspection results of multiple samples within the target defect certification subspace, and this set is the sample defect quality inspection result cluster.

[0047] Furthermore, this application also includes:

[0048] Calculate the first - dimension horizontal distribution quantity and the first - dimension vertical distribution quantity of the first - dimension horizontal starting line and the first - dimension vertical starting line respectively based on the first step length;

[0049] Move the first - dimension horizontal starting line by the first step length in the left - right directions of the line respectively to obtain a set of first - dimension horizontal stage lines, and determine whether the set of first - dimension horizontal stage distribution quantities of the set of first - dimension horizontal stage lines is greater than the first - dimension horizontal distribution quantity. If so, continue to move in the corresponding direction based on the first - dimension horizontal stage line;

[0050] If not, accept moving in the corresponding direction based on the first - dimension horizontal stage line with a certain probability;

[0051] Continue to move in the corresponding direction until the difference between the first - dimension horizontal stage distribution quantities after two adjacent moves is less than a preset gain, and then stop the movement in the corresponding direction to obtain two horizontal boundary lines.

[0052] Specifically, based on the first step length, calculate the distribution quantity of data points on the first - dimension horizontal starting line. This is achieved by counting the number of data points on or near this line. Similarly, calculate the distribution quantity of data points on the first - dimension vertical starting line. Move the first - dimension horizontal starting line by the first step length in the left and right directions respectively to obtain two new lines, forming a set of first - dimension horizontal stage lines. For each new line, calculate the distribution quantity of data points on it. Compare the distribution quantity of data points on the new line with the initial first - dimension horizontal distribution quantity. If the distribution quantity in a certain direction is greater than the initial distribution quantity, continue to move in that direction. If the distribution quantity does not increase, then with a certain probability, which can be a gradually decreasing probability, decide whether to continue moving in that direction. Continuously repeat the above process, recalculate the distribution quantity after each move, and determine the next moving direction based on the change in the distribution quantity. When the difference between the first - dimension horizontal stage distribution quantities after two adjacent moves is less than a preset gain threshold, stop the movement in the corresponding direction. Through the above iterative process, finally, two horizontal boundary lines can be obtained, which respectively represent the effective boundaries of data points in the horizontal direction.

[0053] Furthermore, this application also includes:

[0054] Conduct vertical distribution recognition on the first - dimension vertical starting line to obtain two vertical boundary lines;

[0055] Take the area inside the region constructed by the two horizontal boundary lines and the two vertical boundary lines as the first - dimension defect plane.

[0056] Specifically, move the vertical starting line upward and downward, and recalculate the distribution quantity of data points after each movement. According to the change in the distribution quantity, it can be determined whether to continue moving the vertical starting line upward or downward. When the movement no longer results in a significant increase in the distribution quantity, or reaches a certain preset number of movement times or threshold, stop the movement and determine two vertical boundary lines. Now there are two horizontal boundary lines and two vertical boundary lines, which define a rectangular area. The interior of this rectangular area is the first-dimensional defect plane. The points on this plane represent possible defects in the first dimension, such as the severity of bubble defects, and the second dimension, such as filling uniformity.

[0057] Furthermore, this application also includes:

[0058] Respectively multiply the ratio of each of the K defect degrees to the sum of the K defect degrees by a preset parameter adjustment step length to obtain the K parameter adjustment step lengths;

[0059] Use the K parameter adjustment step lengths to respectively fine-tune the current parameter values of the process nodes in the K associated node sets to obtain K sets of fine-tuned associated node parameter values;

[0060] Based on the N node parameter tolerance bandwidths, perform constraint authentication on the K sets of fine-tuned associated node parameter values. When the authentication passes, obtain the K sets of fine-tuned parameters for the associated nodes.

[0061] Specifically, for each associated node, calculate the ratio of its defect degree to the sum of the defect degrees of all associated nodes. This ratio reflects the contribution degree of this node to the overall defects. Multiply this ratio by the preset parameter adjustment step length to obtain the parameter adjustment step length for each associated node. This step length will be used for subsequent parameter fine-tuning. Use the calculated parameter adjustment step lengths to fine-tune the current parameter values of the process nodes in each associated node set. The fine-tuning direction can be determined according to the actual situation, which may be an increase or a decrease. After fine-tuning, a new set of parameter values is obtained, and these values constitute the set of fine-tuned associated node parameter values. For each node parameter, there is a preset tolerance bandwidth. This bandwidth defines the acceptable change range of the parameter. Compare the fine-tuned parameter values with the tolerance bandwidth to check whether the fine-tuned parameters meet the constraint conditions. That is, check whether the fine-tuned parameters fall within the preset tolerance bandwidth. If all the fine-tuned parameters meet the constraint conditions, then the authentication passes, and a set of fine-tuned parameters for the associated nodes that have been fine-tuned and authenticated will be obtained. When all steps are completed, a set of fine-tuned parameters for the associated nodes that have been optimized and verified is obtained. The parameter values in this set take into account both the defect contribution degree of each node and meet the preset constraint conditions.

[0062] Furthermore, this application also includes:

[0063] When the authentication fails, among the multiple associated node sets where the filling process nodes that fail the authentication are located, the parameter adjustment step size of the associated node set with the largest parameter adjustment step size is restricted until the authentication passes, and the K sets of fine-tuned parameters of the associated nodes are obtained.

[0064] Specifically, for each problem node, determine the associated node set to which it belongs. A node may belong to multiple associated node sets. Among all the associated node sets containing the problem node, find the associated node set with the largest parameter adjustment step size. Restrict its parameter adjustment step size. The restriction ratio can be set according to the actual situation, such as halving or reducing it according to a certain fixed ratio. Use the restricted parameter adjustment step size to re-fine-tune the process node parameters in this associated node set. Re-perform the constraint authentication on the fine-tuned parameter values to check whether all parameters meet the requirements of the tolerance bandwidth. If the constraint authentication still fails, repeat the above steps, and each iteration restricts the associated node set with the largest parameter adjustment step size until the parameters of all nodes meet the constraint conditions. When the parameters of all nodes pass the constraint authentication, record the set of fine-tuned parameters of the associated nodes at this time, which is the final K sets of fine-tuned parameters of the associated nodes.

[0065] In summary, a method for adjusting the filling process of die bonding glue provided by the present application has the following technical effects:

[0066] Node extraction is performed through an interactive die bonding glue filling production line to obtain N filling process nodes. Among them, the N filling process nodes have N node parameter tolerance bandwidths; the quality inspection results of multiple die bonding glue filling samples within a preset feedback window are extracted, and the extracted quality inspection results of multiple samples are defect-certified according to defect characteristics to generate a sample defect quality inspection result cluster. Among them, the quality inspection results of the multiple samples have multiple defect feature sets; using the defect feature type as an index, defect degree analysis is performed on the multiple defect feature sets in the sample defect quality inspection result cluster and the calibrated defect feature value set to obtain K defect degrees of K defect features; the defect feature-node adaptation network layer is called to perform node recognition on the K defect features to obtain K associated node sets; K parameter adjustment step sizes are respectively matched according to the K defect degrees, and based on the K parameter adjustment step sizes, the current parameter values of the process nodes in the K associated node sets are finely adjusted with the N node parameter tolerance bandwidths as constraints to obtain K associated node fine-tuning parameter sets; the process parameters of the die bonding glue filling production line are adjusted according to the K associated node fine-tuning parameter sets, effectively solving the technical problems in the prior art that the process parameter settings of the die bonding glue filling production line are unreasonable and lack a dynamic adjustment mechanism, resulting in unstable product quality, easy occurrence of defects, and further affecting production efficiency and product yield. The quality control and process parameter adjustment process of the die bonding glue filling production line are optimized, and the product quality and production efficiency are comprehensively improved.

[0067] Embodiment 2

[0068] Based on the filling process adjustment method of a die bonding glue in the foregoing embodiment and the same inventive concept, the present application further provides a filling process adjustment system for a die bonding glue. Please refer to the appendix Figure 2 , the system includes:

[0069] A node extraction module 11, which is used to perform node extraction through an interactive die bonding glue filling production line to obtain N filling process nodes. Among them, the N filling process nodes have N node parameter tolerance bandwidths;

[0070] A defect certification module 12, which is used to extract the quality inspection results of multiple die bonding glue filling samples within a preset feedback window, defect-certify the extracted quality inspection results of multiple samples according to defect characteristics, and generate a sample defect quality inspection result cluster. Among them, the quality inspection results of the multiple samples have multiple defect feature sets;

[0071] A defect degree acquisition module 13, which is used to perform defect degree analysis on the multiple defect feature sets in the sample defect quality inspection result cluster and the calibrated defect feature value set with the defect feature type as an index to obtain K defect degrees of K defect features;

[0072] A node recognition module 14, which is used to call the defect feature-node adaptation network layer to recognize nodes for the K defect features, and obtain K associated node sets;

[0073] A parameter set acquisition module 15, which is used to respectively match K parameter adjustment steps according to the K defect degrees, and based on the N node parameter tolerance bandwidths, fine-tune the current parameter values of the process nodes in the K associated node sets according to the K parameter adjustment steps, and obtain K associated node fine-tuning parameter sets;

[0074] A parameter adjustment module 16, which is used to adjust the process parameters of the die bonding glue filling production line according to the K associated node fine-tuning parameter sets.

[0075] Furthermore, the defect certification module 12 in the system is also used for:

[0076] Construct a defect certification space based on a preset defect feature set, where the preset defect feature set includes bubble defects, filling uniformity, and edge integrity;

[0077] Construct multiple spatial coordinate points with the defect feature types and defect values of the multiple defect feature sets, and mark the multiple spatial coordinate points with the multiple sample quality inspection results;

[0078] Perform defect certification on the multiple die bonding glue filling samples based on the multiple spatial coordinate points and the defect certification space, and obtain a sample defect quality inspection result cluster.

[0079] Furthermore, the system also includes a quality result cluster acquisition module, and the quality result cluster acquisition module is used for:

[0080] Map the multiple spatial coordinate points in the defect certification space to the first-dimensional plane to obtain a first-dimensional scatter plot, where the first-dimensional plane is a plane formed by the first-dimensional coordinate axis and the second-dimensional coordinate axis of the defect certification space;

[0081] Extract the upper and lower boundaries of the first-dimensional scatter plot, and use the straight line passing through the center of the upper and lower boundaries and parallel to the first-dimensional coordinate axis as the first-dimensional horizontal starting line;

[0082] Extract the left and right boundaries of the first-dimensional scatter plot, and use the straight line passing through the center of the upper and lower boundaries and perpendicular to the first-dimensional coordinate axis as the first-dimensional vertical starting line;

[0083] Perform horizontal distribution recognition and vertical distribution recognition respectively based on the first - dimension horizontal starting line and the first - dimension vertical starting line to obtain a first - dimension defective plane;

[0084] Map multiple spatial coordinate points in the defect authentication space to a second - dimension plane to obtain a second - dimension scatter plot, and perform defect authentication analysis on the second - dimension scatter plot to obtain a second - dimension defective plane, where the second - dimension plane is the plane formed by the second - dimension coordinate axis and the third - dimension coordinate axis of the defect authentication space;

[0085] Map multiple spatial coordinate points in the defect authentication space to a third - dimension plane to obtain a third - dimension scatter plot, and perform defect authentication analysis on the third - dimension scatter plot to obtain a third - dimension defective plane, where the third - dimension plane is the plane formed by the first - dimension coordinate axis and the third - dimension coordinate axis of the defect authentication space;

[0086] Construct a target defect authentication subspace based on the first - dimension defective plane, the second - dimension defective plane, and the third - dimension defective plane, and use the multiple sample quality detection results corresponding to the multiple spatial coordinate points in the target defect authentication subspace as a sample defect quality detection result cluster.

[0087] Furthermore, the system further includes a boundary line acquisition module, and the boundary line acquisition module is used for:

[0088] Calculate the first - dimension horizontal distribution quantity and the first - dimension vertical distribution quantity of the first - dimension horizontal starting line and the first - dimension vertical starting line respectively based on the first step size;

[0089] Move the first - dimension horizontal starting line by the first step size in the left - right directions of the line respectively to obtain a set of first - dimension horizontal stage lines, and determine whether the set of first - dimension horizontal stage distribution quantities of the set of first - dimension horizontal stage lines is greater than the first - dimension horizontal distribution quantity. If so, continue to move in the corresponding direction based on the first - dimension horizontal stage line;

[0090] If not, accept moving in the corresponding direction based on the first - dimension horizontal stage line with a certain probability;

[0091] Stop moving in the corresponding direction until the difference between the first - dimension horizontal stage distribution quantities after two adjacent moves is less than a preset gain, and obtain two horizontal boundary lines.

[0092] Furthermore, the system further includes a first - dimension defective plane acquisition module, and the first - dimension defective plane acquisition module is used for:

[0093] Perform vertical distribution recognition on the first - dimension vertical starting line to obtain two vertical boundary lines;

[0094] The interior of the region formed by the two horizontal boundary lines and the two vertical boundary lines is taken as the first-dimensional defect plane.

[0095] Further, the system further includes a parameter set acquisition module, and the parameter set acquisition module is used for:

[0096] Multiply the ratio of each of the K defect degrees to the sum of the K defect degrees by a preset parameter adjustment step length to obtain the K parameter adjustment step lengths;

[0097] Use the K parameter adjustment step lengths to slightly adjust the current parameter values of the process nodes in the K associated node sets respectively to obtain a set of K fine-tuned associated node parameter values;

[0098] Based on the N node parameter tolerance bandwidths, perform constraint authentication on the set of K fine-tuned associated node parameter values. When the authentication passes, obtain the set of K associated node fine-tuned parameters.

[0099] Further, the system further includes a compensation and reduction module, and the compensation and reduction module is used for:

[0100] When the authentication fails, reduce the parameter adjustment step length of the associated node set with the largest parameter adjustment step length among the multiple associated node sets where the filling process node that fails is located until the authentication passes, and obtain the set of K associated node fine-tuned parameters.

[0101] In the present specification, the various embodiments are described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The Figure 1 A method for adjusting the filling process of die bonding glue and a specific example in the first embodiment are equally applicable to the system for adjusting the filling process of die bonding glue in this embodiment. Through the detailed description of the method for adjusting the filling process of die bonding glue above, those skilled in the art can clearly know the system for adjusting the filling process of die bonding glue in this embodiment. Therefore, for the sake of simplicity of the specification, it will not be described in detail here. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple. For related parts, refer to the description in the method part.

[0102] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0103] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of this application and its equivalent technologies, this application also intends to include these modifications and variations.

Claims

1. A method for adjusting the filling process of a die-bonding adhesive, characterized in that: The method is applied to a filling process adjustment system, the filling process adjustment system is connected to a die-bonding adhesive filling production line in communication, and the method includes: The interactive die-bonding adhesive filling production line performs node extraction to obtain N filling process nodes, wherein the N filling process nodes have N node parameter tolerance bandwidths; Extracting quality inspection results of multiple die-bonding adhesive filling samples within a preset feedback window, performing defect authentication on the extracted multiple sample quality inspection results according to defect characteristics, and generating a sample defect quality inspection result cluster, wherein the multiple sample quality inspection results have multiple defect feature sets; Taking the defect feature type as an index, performing defect degree analysis on multiple defect feature sets and calibrated defect feature value sets in the sample defect quality detection result cluster to obtain K defect degrees of the K defect features; Calling the defect feature-node adaptation network layer to perform node identification on the K defect features to obtain K associated node sets; Matching K parameter adjustment step sizes according to the K defect degrees respectively, taking the N node parameter tolerance bandwidths as constraints, and fine-tuning current parameter values ​​of process nodes in the K associated node sets based on the K parameter adjustment step sizes to obtain K associated node fine-tuning parameter sets; The process parameters of the die-bonding adhesive filling production line are adjusted according to the K associated node fine-tuning parameter sets.

2. The method according to claim 1, characterized in that Extracting quality inspection results of multiple die-bonding adhesive filling samples within a preset feedback window, performing defect authentication on the extracted multiple sample quality inspection results according to defect characteristics, and generating a sample defect quality inspection result cluster, the method comprising: Constructing a defect authentication space based on a preset defect feature set, wherein the preset defect feature set includes bubble defects, filling uniformity, and edge integrity; Constructing a plurality of spatial coordinate points based on the defect feature types and defect values ​​of the plurality of defect feature sets, and marking the plurality of spatial coordinate points using the plurality of sample quality inspection results; Defect authentication is performed on the plurality of die-bonding adhesive filling samples based on the plurality of spatial coordinate points and the defect authentication space to obtain a sample defect quality detection result cluster.

3. The method according to claim 2, characterized in that Based on the multiple spatial coordinate points and the defect authentication space, defect authentication is performed on the multiple die-bonding adhesive filling samples to obtain a sample defect quality detection result cluster, the method comprising: Mapping a plurality of spatial coordinate points in the defect authentication space to a first-dimensional plane to obtain a first-dimensional scatter plot, wherein the first-dimensional plane is a plane formed by a first-dimensional coordinate axis and a second-dimensional coordinate axis of the defect authentication space; Extracting the upper and lower boundaries of the first-dimensional scatter plot, and taking a straight line passing through the centers of the upper and lower boundaries and being parallel to the first-dimensional coordinate axis as a first-dimensional horizontal starting straight line; Extracting the left and right boundaries of the first-dimensional scatter plot, and taking a straight line passing through the centers of the upper and lower boundaries and being perpendicular to the first-dimensional coordinate axis as a first-dimensional longitudinal starting straight line; Based on the first-dimension transverse starting straight line and the first-dimension longitudinal starting straight line, transverse distribution recognition and longitudinal distribution recognition are performed respectively to obtain a first-dimension defect plane; Mapping a plurality of spatial coordinate points in the defect authentication space to a second-dimensional plane to obtain a second-dimensional scatter plot, performing defect authentication analysis on the second-dimensional scatter plot to obtain a second-dimensional defect plane, wherein the second-dimensional plane is a plane formed by the second-dimensional coordinate axis and the third-dimensional coordinate axis of the defect authentication space; Mapping a plurality of spatial coordinate points in the defect authentication space to a third-dimensional plane to obtain a third-dimensional scatter plot, performing defect authentication analysis on the third-dimensional scatter plot to obtain a third-dimensional defect plane, wherein the third-dimensional plane is a plane formed by the first-dimensional coordinate axis and the third-dimensional coordinate axis of the defect authentication space; A target defect authentication subspace is constructed based on the first dimensional defect plane, the second dimensional defect plane and the third dimensional defect plane, and multiple sample quality detection results corresponding to multiple spatial coordinate points in the target defect authentication subspace are used as a sample defect quality detection result cluster.

4. The method according to claim 3, characterized in that Based on the first-dimension transverse starting straight line and the first-dimension longitudinal starting straight line, transverse distribution identification and longitudinal distribution identification are performed respectively to obtain a first-dimension defect plane, the method comprising: Based on the first step length, respectively calculate the first-dimensional lateral distribution amount and the first-dimensional longitudinal distribution amount of the first-dimensional lateral starting straight line and the first-dimensional longitudinal starting straight line; Move the first-dimensional horizontal starting straight line to the left and right directions of the straight line by a first step length respectively, obtain a first-dimensional horizontal stage straight line set, determine whether the first-dimensional horizontal stage distribution amount set of the first-dimensional horizontal stage straight line set is greater than the first-dimensional horizontal distribution amount, and if so, continue to move in the corresponding direction based on the first-dimensional horizontal stage straight line; If not, then accepting the horizontal phase straight line based on the first dimension to continue moving in the corresponding direction according to a certain probability; Until the difference in the first-dimensional lateral stage distribution amount after two adjacent movements is less than the preset gain, the movement in the corresponding direction is stopped to obtain two lateral boundary lines.

5. The method according to claim 4, characterized in that After obtaining two lateral boundary lines of the first-dimensional defect plane, the method comprises: Perform longitudinal distribution recognition on the longitudinal starting straight line of the first dimension to obtain two longitudinal boundary straight lines; The interior of the area constructed by the two transverse boundary lines and the two longitudinal boundary lines is used as the first-dimensional defect plane.

6. The method according to claim 3, characterized in that Matching K parameter adjustment step sizes according to the K defect degrees respectively, taking the N node parameter tolerance bandwidths as constraints, and fine-tuning the current parameter values ​​of the process nodes in the K associated node sets based on the K parameter adjustment step sizes to obtain the K associated node fine-tuning parameter sets, the method comprising: Multiplying the ratio of the K defect degrees to the sum of the K defect degrees by the preset parameter adjustment step length to obtain the K parameter adjustment step lengths; Using the K parameter adjustment step sizes, respectively fine-tune the current parameter values ​​of the process nodes in the K associated node sets to obtain K fine-tuned associated node parameter value sets; Constrained authentication is performed on the K fine-tuning associated node parameter value sets based on the N node parameter tolerance bandwidths, and when the authentication passes, the K associated node fine-tuning parameter sets are obtained.

7. The method according to claim 6, characterized in that The method comprises: When the authentication fails, the parameter adjustment step of the associated node set with the largest parameter adjustment step among the multiple associated node sets where the failed filling process nodes are located is limited until the authentication passes, and the K associated node fine-tuning parameter sets are obtained.

8. A filling process adjustment system for a die-bonding adhesive, characterized in that: For implementing the steps of the method according to any one of claims 1 to 7, the system comprises: A node extraction module, wherein the node extraction module is used to perform node extraction on an interactive die-bonding adhesive filling production line to obtain N filling process nodes, wherein the N filling process nodes have N node parameter tolerance bandwidths; A defect authentication module, the defect authentication module is used to extract quality inspection results of multiple die-bonding adhesive filling samples within a preset feedback window, perform defect authentication on the extracted multiple sample quality inspection results according to defect characteristics, and generate a sample defect quality inspection result cluster, wherein the multiple sample quality inspection results have multiple defect feature sets; A defect degree acquisition module, wherein the defect degree acquisition module is used to perform defect degree analysis on multiple defect feature sets and calibrated defect feature value sets in the sample defect quality detection result cluster by taking defect feature type as an index, and obtain K defect degrees of K defect features; A node identification module, the node identification module is used to call the defect feature-node adaptation network layer to perform node identification on the K defect features to obtain K associated node sets; A parameter set acquisition module, the parameter set acquisition module is used to match K parameter adjustment step sizes according to the K defect degrees, respectively, take the N node parameter tolerance bandwidths as constraints, and fine-tune the current parameter values ​​of the process nodes in the K associated node sets based on the K parameter adjustment step sizes to obtain the K associated node fine-tuning parameter sets; A parameter adjustment module is used to adjust the process parameters of the die-bonding adhesive filling production line according to the K associated node fine-tuning parameter sets.

Images