Method and system for reducing packaging adsorption defects based on AI intelligence

Abstract

The invention relates to the technical field of artificial intelligence, and discloses a method for reducing packaging adsorption defects based on AI intelligence, and the method comprises the steps: collecting multi-source data between a to-be-packaged element and a packaging film; performing time-space alignment on the multi-source data, performing data fusion on the aligned multi-source data, analyzing the adsorption force between the to-be-packaged element and the packaging film, determining the adsorption force type of the adsorption force, and constructing an adsorption force evolution network between the to-be-packaged element and the packaging film so as to calculate the adsorption defect probability between the to-be-packaged element and the packaging film; calculating pulse parameters of the pulse generator, and calculating a dynamic diffraction coefficient of the pulse generator so as to determine pulse wavefront phase distribution and a trigger time sequence of the pulse generator; and determining a pulse optimization scheme of the to-be-packaged element and the packaging film in the packaging process to execute packaging of the to-be-packaged element to obtain a target packaging element. According to the invention, the quality and efficiency of products in the packaging process can be improved.

Description

Technical Field

[0001] The present invention relates to a method for reducing packaging adsorption defects based on AI intelligence, and belongs to the technical field of artificial intelligence. Background Art

[0002] Package adsorption defects occur during the packaging process due to the adsorption of unwanted particles, gas molecules, moisture, or other contaminants onto the material surface, resulting in a decrease in package quality. Reducing package adsorption defects can avoid operational interruptions, retry attempts, and equipment cleanup caused by adsorption issues, shortening production cycles and improving overall production line efficiency.

[0003] Traditional methods for reducing packaging adsorption defects rely primarily on experience, physical principles, and standardized process control. This approach, based on physical and chemical principles and long-term accumulated experience, controls external conditions (environment, materials, equipment) to reduce or avoid adsorption. However, due to the difficulty in coping with complex and changing production environments and the challenges posed by new materials / processes, the defect rate fluctuates greatly, resulting in limited and unstable effects in reducing packaging adsorption defects. Summary of the Invention

[0004] The present invention provides an AI-based intelligent method for reducing packaging adsorption defects, the main purpose of which is to improve the quality and efficiency of products during the packaging process.

[0005] To achieve the above objectives, the present invention provides an AI-based intelligent method for reducing packaging adsorption defects, comprising: Real-time collection of multi-source data between the components to be packaged and the packaging film; Performing spatiotemporal alignment on the multi-source data to obtain aligned multi-source data, and performing data fusion on the aligned multi-source data to obtain a comprehensive data set; analyzing, based on the comprehensive data set, the adsorption force between the component to be packaged and the packaging film using a defect analysis module of a preset AI adsorption defect analysis model to determine an adsorption force type of the adsorption force; and constructing, based on the adsorption force and the adsorption force type, an adsorption force evolution network between the component to be packaged and the packaging film using a network evolution module of the AI ​​adsorption defect analysis model to calculate a probability of adsorption defects between the component to be packaged and the packaging film; Analyzing the factors influencing the adsorption defect probability, configuring a pulse generator for the component to be packaged and the packaging film during the packaging process, and calculating pulse parameters of the pulse generator based on the influencing factors, wherein the pulse parameters include: injection position, pulse pressure, and pulse duration. Combining the pulse parameters with the comprehensive data set, calculating the dynamic diffraction coefficient of the pulse generator to determine the pulse wavefront phase distribution and trigger timing of the pulse generator; In combination with the pulse wavefront phase distribution and the trigger timing, the decision module of the AI ​​adsorption defect analysis model is used to determine the pulse optimization scheme of the component to be packaged and the packaging film during the packaging process to perform the packaging of the component to be packaged and obtain the target packaged component.

[0006] Optionally, the real-time collection of multi-source data between the component to be packaged and the packaging film includes: Clarify the data collection target required between the component to be packaged and the packaging film; Based on the data acquisition target, configuring multiple sensors between the component to be packaged and the packaging film; Determining a connection method and network layout of the multiple sensors; constructing a sensor network of the multiple sensors according to the connection mode and the network layout; Determining a sampling rate of the multi-sensor to construct a time synchronization network of the multi-sensor; Multi-source data between the component to be packaged and the packaging film is collected according to the sensor network and the time synchronization network.

[0007] Optionally, performing spatiotemporal alignment on the multi-source data to obtain aligned multi-source data includes: Preprocessing the multi-source data to obtain preprocessed multi-source data; Determining a main time axis of the pre-processed multi-source data and identifying a timestamp of the pre-processed multi-source data; performing time alignment on the preprocessed multi-source data according to the main time axis and the timestamp to obtain time-aligned multi-source data; Constructing a global reference coordinate system for the time-aligned multi-source data and extracting local coordinate information of the time-aligned multi-source data; Converting the local coordinate information into the global reference coordinate system to obtain global coordinates; Based on the global coordinates, spatial registration is performed on the time-aligned multi-source data to obtain aligned multi-source data.

[0008] Optionally, analyzing the adsorption force between the component to be packaged and the packaging film by using a defect analysis module of a preset AI adsorption defect analysis model based on the comprehensive data set includes: Extracting key characteristic parameters of the comprehensive data set, wherein the key characteristic parameters include: material properties, process parameters, surface parameters, and environmental parameters; Based on the key characteristic parameters, the adsorption force between the component to be encapsulated and the encapsulation film is calculated using a hybrid adsorption force algorithm in the defect analysis module, wherein the hybrid adsorption force algorithm includes: ; in, Indicates adsorption force, Indicates the strength of molecular interaction between materials in material properties, Indicates the surface distance correction factor between the component to be packaged and the packaging film, represents pi, Indicates the average gap distance between the component to be packaged and the packaging film, represents the equivalent contact area correction factor, Indicates the actual contact area radius in the process parameters, represents the cosine function, Represents the contact angle in the process parameters, represents the surface energy among the surface parameters, Indicates The exponential function with base , represents the fitting parameters, Indicates the relative humidity in the environmental parameters. represents the elastic deformation coefficient, represents the Young's modulus of the film, Indicates the peak-to-valley height of the surface.

[0009] Optionally, constructing an adsorption force evolution network between the component to be encapsulated and the encapsulation film by using a network evolution module of the AI ​​adsorption defect analysis model according to the adsorption force and the adsorption force type includes: dividing the network nodes between the components to be packaged and the packaging film; constructing a node feature vector of the network node according to the adsorption force and the adsorption force type; defining edge connection rules for the network nodes; Determining the adsorption force transmission strength between the network nodes according to the edge connection rule; Based on the node feature vector and the adsorption force transmission strength, a time evolution algorithm of the network node is defined, wherein the time evolution algorithm includes: ; in, Represents a network node In time The node feature vector of represents the activation function, represents the physical constraint matrix, Represents a network node In time The node feature vector of Represents a network node The neighbor set of represents the bias term, represents the relaxation time, Represents a network node The node feature vector of The instantaneous rate of change of Determining the instantaneous rate of change of the node feature vector according to the time evolution algorithm; According to the instantaneous change rate, an adsorption force evolution network between the component to be packaged and the packaging film is constructed.

[0010] Optionally, the calculating the probability of adsorption defects between the component to be packaged and the packaging film includes: Analyzing the change rate of the node adsorption force in the corresponding adsorption force evolution network between the component to be encapsulated and the encapsulation film; Determining attention weights between corresponding network nodes of the adsorption force evolution network; Calculating a node adsorption defect probability of the network node according to the attention weight and the node adsorption force change rate; Identifying a boundary of the component to be packaged, and calculating a normalized distance between the network node and the boundary; The adsorption defect probability between the component to be packaged and the packaging film is calculated according to the normalized distance and the node adsorption defect probability.

[0011] Optionally, calculating the pulse parameters of the pulse generator based on the influencing factor includes: Fitting a pulse parameter analysis model of the pulse generator based on the influencing factors; Determining the constraints and objective function of the pulse parameter analysis model; constructing an iterative mechanism for the pulse parameter analysis model according to the constraint conditions and the objective function; determining initial pulse parameters of the pulse parameter analysis model, and analyzing convergence of the initial pulse parameters; When the convergence does not meet a preset convergence threshold, optimizing the initial pulse parameters through the iterative mechanism to obtain optimized pulse parameters; Analyzing the post-optimization convergence of the optimized pulse parameters; When the convergence after optimization meets the convergence threshold, the optimized pulse parameters are used as the pulse parameters of the pulse generator.

[0012] Optionally, the calculating the dynamic diffraction coefficient of the pulse generator by combining the pulse parameters and the comprehensive data set includes: extracting comprehensive data features of the comprehensive data set; Performing feature normalization fusion on the comprehensive data features and the pulse parameters to obtain normalized fusion features; Calculating the fluctuation field value of the pulse generator according to the normalized fusion feature; Calculating the defect area energy and the total incident energy of the pulse generator according to the fluctuation field value; The dynamic diffraction coefficient of the pulse generator is calculated based on the defect area energy and the incident total energy.

[0013] Optionally, determining the pulse wavefront phase distribution and trigger timing of the pulse generator includes: determining a pulse wavelength of the pulse generator; Identifying a defect position corresponding to the pulse generator and analyzing a defect height difference at the defect position; Determining the pulse wavefront phase distribution of the pulse generator according to the pulse wavelength, the defect height difference and the dynamic diffraction coefficient corresponding to the pulse generator; constructing a time response curve of the dynamic diffraction coefficient; Calculating the delay time of the pulse corresponding to the pulse generator based on the time response curve; The triggering timing of the pulse generator is determined according to the delay time.

[0014] In order to solve the above problems, the present invention also provides an AI-based intelligent system for reducing package adsorption defects, the system comprising: Multi-source data acquisition module, used to collect multi-source data between the components to be packaged and the packaging film in real time; A multi-source data fusion module is used to perform spatiotemporal alignment on the multi-source data to obtain aligned multi-source data, and perform data fusion on the aligned multi-source data to obtain a comprehensive data set; an adsorption defect probability calculation module, configured to analyze, based on the comprehensive data set and using a defect analysis module of a preset AI adsorption defect analysis model, the adsorption force between the component to be packaged and the packaging film, determine the adsorption force type of the adsorption force, and construct, based on the adsorption force and the adsorption force type, an adsorption force evolution network between the component to be packaged and the packaging film using a network evolution module of the AI ​​adsorption defect analysis model to calculate the adsorption defect probability between the component to be packaged and the packaging film; a pulse parameter analysis module, configured to analyze factors influencing the adsorption defect probability, configure a pulse generator for the component to be packaged and the packaging film during the packaging process, and calculate pulse parameters of the pulse generator based on the influencing factors, wherein the pulse parameters include: injection position, pulse pressure, and pulse duration; and calculate the dynamic diffraction coefficient of the pulse generator by combining the pulse parameters and the comprehensive data set to determine the pulse wavefront phase distribution and trigger timing of the pulse generator; A packaging optimization module is used to combine the pulse wavefront phase distribution and the trigger timing, and determine the pulse optimization scheme of the component to be packaged and the packaging film during the packaging process through the decision module of the AI ​​adsorption defect analysis model, so as to perform the packaging of the component to be packaged and obtain the target packaged component.

[0015] Compared with the problems described in the background technology, the embodiment of the present invention can achieve true process visibility and transparency by real-time collection of multi-source data between the component to be packaged and the packaging film, thereby improving the timeliness and accuracy of defect detection; optionally, the embodiment of the present invention can fuse the aligned multi-source data to obtain a comprehensive data set, which can complement each other and make up for each other's shortcomings, thereby obtaining a more accurate and robust description than a single sensor, covering a wider space, and filling in the regional information that a single sensor cannot observe; the embodiment of the present invention constructs an adsorption force evolution network between the component to be packaged and the packaging film according to the adsorption force and the adsorption force type through the network evolution module of the AI ​​adsorption defect analysis model, which can simulate the dynamic process of the adsorption force size, distribution and different types of adsorption force changing with environmental factors such as time, temperature, pressure, humidity, etc. from the beginning of contact to the completion of packaging and even the subsequent possible use process, so as to analyze the evolution of the adsorption force under specific process parameters or environmental conditions, as well as the critical point of this evolution, thereby providing Early warning of potential defects; by calculating the probability of adsorption defects between the component to be packaged and the packaging film, embodiments of the present invention can achieve accurate defect prediction and risk quantification, thereby significantly reducing defective products caused by adsorption defects in the final product; by determining the pulse wavefront phase distribution and trigger timing of the pulse generator, embodiments of the present invention can ensure that energy is more concentrated on the target area (e.g., the defect point), improving processing efficiency and effectiveness, reducing damage to surrounding areas, and precisely controlling the processes of energy accumulation, thermal effects, and chemical reactions, thereby reducing random fluctuations in pulse characteristics; finally, by combining the pulse wavefront phase distribution and trigger timing, embodiments of the present invention determine the pulse optimization scheme for the component to be packaged and the packaging film during the packaging process through the decision module of the AI ​​adsorption defect analysis model, ensuring more uniform and precise deposition of laser energy in the contact area between the component to be packaged and the packaging film, better adapting to variations in adsorption force, reducing adsorption defects, and thus improving yield consistency between and within batches. Therefore, the AI-based method for reducing packaging adsorption defects provided by embodiments of the present invention can improve product quality and efficiency during the packaging process. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic flow chart of a method for reducing package adsorption defects based on AI intelligence provided by one embodiment of the present invention; Figure 2 A schematic diagram of a module for implementing the AI-based intelligent method for reducing packaging adsorption defects provided in one embodiment of the present invention.

[0017] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0018] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0019] The embodiments of the present application provide an AI-based method for reducing package adsorption defects. The execution entity of the AI-based method for reducing package adsorption defects includes, but is not limited to, at least one of electronic devices such as a server and a terminal that can be configured to execute the method provided by the embodiments of the present application. In other words, the AI-based method for reducing package adsorption defects can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster.

[0020] Example 1: Reference Figure 1 FIG. 1 is a flow chart of a method for reducing package adsorption defects based on AI intelligence according to an embodiment of the present invention. In this embodiment, the method for reducing package adsorption defects based on AI intelligence includes: S1. Real-time collection of multi-source data between the components to be packaged and the packaging film.

[0021] By collecting multi-source data between the encapsulated component and the encapsulation film in real time, embodiments of the present invention can achieve true process visibility and transparency, improving the timeliness and accuracy of defect detection. Multi-source data refers to a collection of data with different information dimensions, such as visual data, pressure data, position data, and environmental data, collected from different types of sensors or data acquisition systems during the thin-film encapsulation process to monitor and analyze the interaction between the encapsulated component and the encapsulation film (particularly adsorption phenomena) in real time.

[0022] As an embodiment of the present invention, the real-time collection of multi-source data between the component to be packaged and the packaging film includes: Clarify the data collection target required between the component to be packaged and the packaging film; Based on the data acquisition target, configuring multiple sensors between the component to be packaged and the packaging film; Determining a connection method and network layout of the multiple sensors; constructing a sensor network of the multiple sensors according to the connection mode and the network layout; Determining a sampling rate of the multi-sensor to construct a time synchronization network of the multi-sensor; Multi-source data between the component to be packaged and the packaging film is collected according to the sensor network and the time synchronization network.

[0023] The data acquisition objective refers to specific information about the interaction area and its surrounding environment, acquired through a sensor system, in order to understand, analyze, and ultimately reduce the adsorption forces generated between the encapsulated component and the encapsulation film during the thin-film encapsulation process. Multi-sensor technology refers to the use of multiple sensors of different types and functions, such as visual sensors (cameras), force / pressure sensors, and displacement / distance sensors, to collect various data (e.g., distance, force, surface condition, position, temperature, etc.) in real time during the thin-film encapsulation process. The connection method refers to the set of physical connection structures, electrical interface standards, communication protocol specifications, and wiring layout schemes used to enable effective communication and data transmission between sensors and data acquisition / processing units. The network layout refers to the physical and logical planning and design of the connection paths between the various nodes involved in the connection method (e.g., sensors, data acquisition units, controllers, processing units, etc.) to ensure efficient data transmission and stable system operation. A sensor network refers to a system consisting of multiple sensor nodes interconnected by a specific connection method (physical and logical connections) and network layout (physical location and communication topology) that can work together to monitor the interaction between components and the film during the thin-film encapsulation process. The sampling rate refers to the number of times a continuous signal (such as analog voltage, current, sound waveform, light intensity, etc.) is sampled per unit time. The time synchronization network refers to a technical system that ensures that the clocks of all devices in the network remain highly consistent.

[0024] Optionally, the connection mode and network layout of the multiple sensors can be determined by passive positioning and association of wireless signal fingerprints, such as identifying the wireless signals of the multiple sensors, analyzing the signal parameters of the wireless signals, and determining the node identity, physical location and communication relationship of the multiple sensors through the signal parameters to determine the connection mode and network layout of the multiple sensors.

[0025] Optionally, the multi-sensor sensor network can be constructed through self-organizing and self-configuring network technologies, such as advanced Mesh network protocols, AI-driven network optimization algorithms, etc.

[0026] Optionally, the time synchronization network of the multiple sensors can be constructed by a distributed clock synchronization algorithm, such as determining the sampling time deviation of the multiple sensors according to the sampling rate, and calculating the synchronization coefficient of the multiple sensors based on the sampling time deviation through the distributed clock synchronization algorithm to construct the time synchronization network of the multiple sensors.

[0027] S2. Performing spatiotemporal alignment on the multi-source data to obtain aligned multi-source data, and performing data fusion on the aligned multi-source data to obtain a comprehensive data set.

[0028] In the embodiment of the present invention, by performing spatiotemporal alignment on the multi-source data, the aligned multi-source data can be directly compared and analyzed without being misleading due to temporal or spatial misalignment. The aligned multi-source data refers to a multi-source data set that has undergone spatiotemporal alignment.

[0029] As an embodiment of the present invention, performing spatiotemporal alignment on the multi-source data to obtain aligned multi-source data includes: Preprocessing the multi-source data to obtain preprocessed multi-source data; Determining a main time axis of the pre-processed multi-source data and identifying a timestamp of the pre-processed multi-source data; performing time alignment on the preprocessed multi-source data according to the main time axis and the timestamp to obtain time-aligned multi-source data; Constructing a global reference coordinate system for the time-aligned multi-source data and extracting local coordinate information of the time-aligned multi-source data; Converting the local coordinate information into the global reference coordinate system to obtain global coordinates; Based on the global coordinates, spatial registration is performed on the time-aligned multi-source data to obtain aligned multi-source data.

[0030] The pre-processed multi-source data refers to the data set obtained after the original multi-source data has been preliminarily cleaned, sorted and normalized. The main time axis refers to the time series selected as the benchmark during the multi-source data time alignment process. The timestamp refers to the time mark attached to each data sample (or data point). The time-aligned multi-source data refers to the data set obtained by adjusting the multi-source data to a unified time benchmark according to the timestamp. The global reference coordinate system refers to a standard spatial framework that unifies all data points in the spatial dimension. The local coordinate information refers to the spatial position information recorded by each data source in its own inherent, independent coordinate system. The global coordinates refer to the spatial position coordinates in the unified global reference coordinate system obtained after the local coordinate information of each data source (sensor) is converted into the unified global reference coordinate system.

[0031] Optionally, the time-aligned multi-source data can be obtained through a deep learning model, such as a recurrent neural network (RNN), a convolutional neural network (CNN), a Transformer, etc.

[0032] Optionally, the global coordinates may be obtained by analyzing a rotation matrix and a translation matrix of the local coordinate information, and transforming the local coordinate information according to the rotation matrix and the translation matrix to obtain the global coordinates.

[0033] By fusing the aligned multi-source data, the present invention generates a comprehensive dataset that leverages the strengths of each sensor and compensates for its weaknesses. This allows for a more accurate and robust description than a single sensor, covers a wider area, and fills in information about areas that cannot be observed by a single sensor. The comprehensive dataset refers to the data set obtained after spatiotemporal alignment and multi-source data fusion.

[0034] Optionally, the comprehensive dataset can be obtained through a fusion algorithm, such as Scikit-learn, TensorFlow, PyTorch, etc.

[0035] S3. Analyze the adsorption force between the component to be packaged and the packaging film using a defect analysis module of a preset AI adsorption defect analysis model based on the comprehensive data set to determine the adsorption force type of the adsorption force. Construct an adsorption force evolution network between the component to be packaged and the packaging film using a network evolution module of the AI ​​adsorption defect analysis model based on the adsorption force and the adsorption force type to calculate the probability of adsorption defects between the component to be packaged and the packaging film.

[0036] The embodiment of the present invention can more accurately invert and evaluate the magnitude of the adsorption force and its changing trend under actual working conditions by analyzing the adsorption force between the component to be packaged and the packaging film based on the comprehensive data set through the defect analysis module of the preset AI adsorption defect analysis model, thereby realizing early defect warning and root cause analysis. Among them, the preset AI adsorption defect analysis model refers to a trained mathematical model specifically used to analyze adsorption problems and defects in the packaging process. The defect analysis module refers to a specific functional unit in the AI ​​adsorption defect analysis model that is specifically responsible for extracting and analyzing the adsorption state information between the component to be packaged and the packaging film from the comprehensive data set. The adsorption force refers to the mutual attraction or adhesion force generated between the component to be packaged and the packaging film when the two objects are in contact or close to each other.

[0037] As an embodiment of the present invention, analyzing the adsorption force between the component to be packaged and the packaging film using a defect analysis module of a preset AI adsorption defect analysis model based on the comprehensive data set includes: Extracting key characteristic parameters of the comprehensive data set, wherein the key characteristic parameters include: material properties, process parameters, surface parameters, and environmental parameters; Based on the key characteristic parameters, the adsorption force between the component to be encapsulated and the encapsulation film is calculated using a hybrid adsorption force algorithm in the defect analysis module, wherein the hybrid adsorption force algorithm includes: ; in, Indicates adsorption force, Indicates the strength of molecular interaction between materials in material properties, Indicates the surface distance correction factor between the component to be packaged and the packaging film, represents pi, Indicates the average gap distance between the component to be packaged and the packaging film, represents the equivalent contact area correction factor, Indicates the actual contact area radius in the process parameters, represents the cosine function, Represents the contact angle in the process parameters, represents the surface energy among the surface parameters, Indicates The exponential function with base , represents the fitting parameters, Indicates the relative humidity in the environmental parameters. represents the elastic deformation coefficient, represents the Young's modulus of the film, Indicates the peak-to-valley height of the surface.

[0038] The key characteristic parameters refer to the data extracted from the comprehensive dataset that are identified as being most important and influential in analyzing the adsorption force between the component to be encapsulated and the encapsulation film. Material properties refer to the inherent physical and chemical properties of the materials that comprise the component to be encapsulated and the encapsulation film, such as molecular interaction strength, Young's modulus, and elastic deformation coefficient. Process parameters refer to the various conditions, settings, and operating methods that are set, controlled, or actually occur during the encapsulation process, such as encapsulation temperature, encapsulation pressure, contact angle, and contact area radius. Surface parameters describe the physical and chemical properties of the contact interface or adjacent interface between the component to be encapsulated and the encapsulation film, such as surface energy and peak-to-valley height. Environmental parameters refer to the state and conditions of the surrounding environment (i.e., the space where the encapsulation process occurs) during the encapsulation process, such as humidity and cleanliness. The hybrid adsorption force algorithm is a mathematical algorithm that combines multiple physical and chemical adsorption mechanisms to calculate the adsorption force between the component to be encapsulated and the encapsulation film. The molecular interaction strength between materials refers to the magnitude of the mutual attraction or repulsion between the two materials that comprise the component to be encapsulated and the encapsulation film at the microscopic level. The surface spacing correction factor refers to the correlation coefficient between the magnitude of the intermolecular force and the distance between the surfaces of an object. The equivalent contact area correction factor refers to the magnification factor of the actual contact area due to the rough surface. The contact angle refers to the contact angle formed by a droplet on a solid surface. The fitting parameters refer to specific parameters introduced or adjusted in the model to enable the hybrid adsorption force algorithm to more accurately match actual experimental data.

[0039] Optionally, the key feature parameters of the comprehensive dataset can be extracted through integrated learning and feature importance evaluation, such as random forest, XGBoost, LightGBM, etc.

[0040] By determining the adsorption force type, embodiments of the present invention can provide more accurate defect prediction and risk assessment, and a deeper understanding of the defect generation mechanism. The adsorption force type refers to the classification of different physical or chemical mechanisms that contribute to the total adsorption force between the encapsulated component and the encapsulation film, such as van der Waals forces, electrostatic forces, and chemical bonding forces.

[0041] Alternatively, the adsorption force type of the adsorption force can be determined by high-resolution microscopic characterization techniques, such as scanning probe microscopy, spectroscopy, etc.

[0042] The embodiment of the present invention constructs the adsorption force evolution network between the component to be packaged and the packaging film through the network evolution module of the AI ​​adsorption defect analysis model based on the adsorption force and the type of adsorption force, which can simulate the dynamic process of the adsorption force size, distribution and different types of adsorption force changing with environmental factors such as time, temperature, pressure, humidity, etc. from the beginning of contact to the completion of packaging and even the subsequent possible use. It can analyze the evolution of adsorption force under specific process parameters or environmental conditions, as well as the critical point of such evolution, and thus provide early warning of potential defects. Among them, the network evolution module refers to a submodule in the AI ​​adsorption defect analysis model that is specifically used to simulate, characterize and predict the dynamic change process of adsorption force and related interactions between the component to be packaged and the packaging film over time, space or process parameters.

[0043] As an embodiment of the present invention, constructing an adsorption force evolution network between the component to be packaged and the packaging film by using the network evolution module of the AI ​​adsorption defect analysis model according to the adsorption force and the adsorption force type includes: dividing the network nodes between the components to be packaged and the packaging film; constructing a node feature vector of the network node according to the adsorption force and the adsorption force type; defining edge connection rules for the network nodes; Determining the adsorption force transmission strength between the network nodes according to the edge connection rule; Based on the node feature vector and the adsorption force transmission strength, a time evolution algorithm of the network node is defined, wherein the time evolution algorithm includes: ; in, Represents a network node In time The node feature vector of represents the activation function, represents the physical constraint matrix, Represents a network node In time The node feature vector of Represents a network node The neighbor set of represents the bias term, represents the relaxation time, Represents a network node The node feature vector of The instantaneous rate of change of Determining the instantaneous rate of change of the node feature vector according to the time evolution algorithm; According to the instantaneous change rate, an adsorption force evolution network between the component to be packaged and the packaging film is constructed.

[0044] The network nodes represent specific micro-regions obtained by discretizing the surfaces of the packaged components and the packaging film. The node feature vectors provide a set of numerical values ​​describing the state and properties of each network node, including the adsorption force, local surface energy and Young's modulus, surface topography parameters, and environmental parameter characteristics of the region. The edge connection rules are methods for establishing connections (edges) between nodes in the adsorption force evolution network and quantifying the strength of these connections. The adsorption force transfer strength refers to the effective contribution of a node to the adsorption force exerted on adjacent nodes through interactions (such as van der Waals forces, capillary forces, mechanical interlocking, etc.). The time evolution algorithm refers to a set of calculation methods that simulate the dynamic changes of node characteristics over time through differential equations or iterative update rules. The activation function refers to a nonlinear transformation component that constrains and normalizes the dynamic update process of node characteristics to ensure that the algorithm can capture complex physical interactions. The physical constraint matrix refers to a weight matrix that embeds physical prior knowledge such as material properties and environmental parameters into the graph neural network. The bias term refers to a learnable parameter in the linear transformation. The relaxation time refers to the time scale required for an algorithm to return from a non-equilibrium state to its equilibrium state, which can be set to 0.1~1s in this application. The instantaneous change rate refers to a physical concept that describes how fast the node feature vector changes at a specific moment.

[0045] Optionally, the edge connection rules of the network nodes can be determined by analyzing the positional relationship of the network nodes. When the positional relationship is an adjacent contact relationship, the physical connection edges of the network nodes are determined; when the positional relationship is a non-adjacent contact relationship, the force transmission probability of the network nodes is calculated, and based on the force transmission probability, the dynamic interaction edges of the network nodes are determined. According to the physical connection edges and the dynamic interaction edges, the edge connection rules of the network nodes are determined.

[0046] Optionally, the adsorption force transmission strength between the network nodes can be determined by a network propagation model, such as a heat diffusion equation, a variant of the SIR model, etc.

[0047] By calculating the probability of adsorption defects between the component to be packaged and the encapsulation film, embodiments of the present invention can achieve accurate defect prediction and risk quantification, significantly reducing the number of defective products in the final product due to adsorption defects. The adsorption defect probability refers to the likelihood of adsorption-related defects occurring between the component to be packaged and the encapsulation film under specific packaging conditions (e.g., specific component to be packaged, encapsulation film, environmental parameters, process parameters, etc.).

[0048] As an embodiment of the present invention, the calculating of the probability of adsorption defects between the component to be packaged and the packaging film includes: Analyzing the change rate of the node adsorption force in the corresponding adsorption force evolution network between the component to be encapsulated and the encapsulation film; Determining attention weights between corresponding network nodes of the adsorption force evolution network; Calculating a node adsorption defect probability of the network node according to the attention weight and the node adsorption force change rate; Identifying a boundary of the component to be packaged, and calculating a normalized distance between the network node and the boundary; The adsorption defect probability between the component to be packaged and the packaging film is calculated according to the normalized distance and the node adsorption defect probability.

[0049] The node adsorption force change rate refers to the rate at which the adsorption force experienced by a single node in the adsorption force evolution network changes over time. The attention weight quantifies the importance of the connection between two nodes in the adsorption force evolution network. The node adsorption defect probability refers to the probability of an adsorption defect occurring at each node. The boundary refers to the physical boundary of the component to be packaged. The normalized distance refers to the normalized value of the actual distance from the network node to the physical boundary of the component.

[0050] Optionally, the node adsorption defect probability of the network node can be calculated using the following formula: ; in, Represents a network node The probability of node adsorption defect, Indicates The exponential function with base , Represents a network node The neighbor set of Represents a network node and network nodes The attention weights between Network Node and network nodes The change rate of the node adsorption force between Indicates the maximum adsorption capacity of the material.

[0051] Optionally, the probability of adsorption defects between the component to be packaged and the packaging film can be calculated using the following formula: ; in, represents the adsorption defect probability, Indicates The exponential function with base , Represents a network node The probability of node adsorption defect, Represents a network node The normalized distance.

[0052] S4. Analyze the influencing factors of the adsorption defect probability, configure a pulse generator for the component to be packaged and the packaging film during the packaging process, and calculate the pulse parameters of the pulse generator based on the influencing factors, wherein the pulse parameters include: injection position, pulse pressure, and pulse duration. Combine the pulse parameters and the comprehensive data set to calculate the dynamic diffraction coefficient of the pulse generator to determine the pulse wavefront phase distribution and trigger timing of the pulse generator.

[0053] By analyzing the factors influencing the adsorption defect probability, embodiments of the present invention can determine the parameters that have the greatest impact on the defect probability, allowing for more precise setting and adjustment of the pulse generator. These factors refer to various factors that can significantly affect the node adsorption defect probability, such as process parameters and structural factors.

[0054] Optionally, the influencing factors of the adsorption defect probability can be analyzed by a multi-factor analysis method, such as variance analysis, regression analysis, principal component analysis, etc.

[0055] The embodiment of the present invention can precisely control the packaging process and reduce adsorption defects by configuring a pulse generator for the component to be packaged and the packaging film during the packaging process. The pulse generator is a device used to control fluid injection or apply local pressure and energy.

[0056] By calculating the pulse parameters of the pulse generator based on the influencing factors, embodiments of the present invention can more accurately determine the parameters of the pulse generator, helping to ensure precise alignment and uniform contact between the encapsulation film and the component to be encapsulated, thereby improving the consistency and reliability of the package. The pulse parameters refer to the specific set values ​​or operating conditions that control the operation of the pulse generator. The injection position refers to the specific coordinate point in three-dimensional space where the pulse generator (such as a nozzle) performs the injection action. The pulse pressure refers to the instantaneous pressure that drives the fluid injection or applies pressure / energy. The pulse duration refers to the length of time the pulse generator maintains a specific pressure or performs the injection action.

[0057] As an embodiment of the present invention, the calculating of the pulse parameters of the pulse generator based on the influencing factor includes: Fitting a pulse parameter analysis model of the pulse generator based on the influencing factors; Determining the constraints and objective function of the pulse parameter analysis model; constructing an iterative mechanism for the pulse parameter analysis model according to the constraint conditions and the objective function; determining initial pulse parameters of the pulse parameter analysis model, and analyzing convergence of the initial pulse parameters; When the convergence does not meet a preset convergence threshold, optimizing the initial pulse parameters through the iterative mechanism to obtain optimized pulse parameters; Analyzing the post-optimization convergence of the optimized pulse parameters; When the convergence after optimization meets the convergence threshold, the optimized pulse parameters are used as the pulse parameters of the pulse generator.

[0058] The pulse parameter analysis model refers to a computational framework based on multi-physics coupling and data-driven optimization for optimizing key operating parameters of pulse generators in packaging processes. Constraints refer to a series of restrictions, rules, or boundaries that must be met during pulse parameter optimization or calculation. The objective function refers to a mathematical function used to measure and evaluate the performance of different pulse parameter combinations during optimization calculations. Initial pulse parameters refer to a set of pulse generator operating parameter values ​​preset before optimization calculations or parameter adjustments. Convergence refers to the process and result of the pulse parameter optimization process in which the objective function value (defect probability after packaging) gradually approaches a stable, ideal value as the number of iterations increases. The convergence threshold refers to a pre-set quantitative criterion used to determine whether the optimization process has achieved its objective or whether it can be stopped. Optimized pulse parameters refer to the pulse generator parameter values ​​obtained after the optimization process (i.e., adjusting the initial pulse parameters through an iterative mechanism) that meet the preset objective function and constraints and whose convergence meets a preset convergence threshold. Post-optimization convergence refers to the characteristic that the optimization results obtained after the pulse parameter optimization process reach a stable state.

[0059] Optionally, the pulse parameter analysis model of the pulse generator can be fitted by multiple linear regression.

[0060] Optionally, the objective function of the pulse parameter analysis model can be determined by a deep reinforcement learning method, such as the Actor-Critic method, PPO, A2C / A3C, DDPG / TD3, etc.

[0061] By combining the pulse parameters with the comprehensive dataset, embodiments of the present invention calculate the dynamic diffraction coefficient of the pulse generator, thereby improving the reliability and consistency of packaging or other application processes and reducing defects caused by pulse output fluctuations. The dynamic diffraction coefficient is a dimensionless parameter that quantifies the effective utilization of pulse energy at the interface of the packaging material.

[0062] As an embodiment of the present invention, the step of calculating the dynamic diffraction coefficient of the pulse generator by combining the pulse parameters and the comprehensive data set includes: extracting comprehensive data features of the comprehensive data set; Performing feature normalization fusion on the comprehensive data features and the pulse parameters to obtain normalized fusion features; Calculating the fluctuation field value of the pulse generator according to the normalized fusion feature; Calculating the defect area energy and the total incident energy of the pulse generator according to the fluctuation field value; The dynamic diffraction coefficient of the pulse generator is calculated based on the defect area energy and the incident total energy.

[0063] Among them, the comprehensive data features refer to the feature set extracted from the comprehensive data set that can represent the core information or key attributes of the data set. The normalized fusion features refer to normalizing the original comprehensive data features and pulse parameters separately or uniformly in the numerical range to eliminate the dimension effect, and then combining (fusing) the normalized comprehensive data features and pulse parameters together to form a new set of features with unified scale and containing multiple aspects of information. The wave field value refers to a quantitative indicator that describes the distribution state and manifestation of the wave generated by the pulse under specific conditions. The defect area energy refers to the total energy of the wave generated by the pulse generator that is transmitted to the specific defect area in the material. The total incident energy refers to the total energy emitted by the pulse generator and acting on the surface of the target material during a single pulse action.

[0064] Optionally, the wave field value of the pulse generator can be calculated by a numerical simulation method, such as a finite difference time domain method (FDTD), a finite element method (FEM), and the like.

[0065] Optionally, the defect area energy and the incident total energy of the pulse generator can be calculated by a numerical integration method, such as a rectangular method, a trapezoidal method, a Simpson method, etc.

[0066] By determining the pulse wavefront phase distribution and trigger timing of the pulse generator, embodiments of the present invention ensure that energy is more concentrated on the target area (e.g., a defect), improving processing efficiency and effectiveness while minimizing damage to surrounding areas. Furthermore, they precisely control the progression of energy accumulation, thermal effects, and chemical reactions, thereby reducing random fluctuations in pulse characteristics. The pulse wavefront phase distribution refers to the spatial distribution of phase values ​​at different locations on a pulse's wavefront at a specific moment. The trigger timing controls when the pulse generator generates pulses, as well as the temporal relationship and precise timing between these pulses (or between pulses and other events).

[0067] As an embodiment of the present invention, determining the pulse wavefront phase distribution and trigger timing of the pulse generator includes: determining a pulse wavelength of the pulse generator; Identifying a defect position corresponding to the pulse generator and analyzing a defect height difference at the defect position; Determining the pulse wavefront phase distribution of the pulse generator according to the pulse wavelength, the defect height difference and the dynamic diffraction coefficient corresponding to the pulse generator; constructing a time response curve of the dynamic diffraction coefficient; Calculating the delay time of the pulse corresponding to the pulse generator based on the time response curve; The triggering timing of the pulse generator is determined according to the delay time.

[0068] The pulse wavelength refers to the central wavelength of the pulse beam output by the pulse generator. The defect location refers to the location where an adsorption defect occurs during the packaging process of the component to be packaged. The defect height difference refers to the vertical height difference between the defect area and the surrounding normal surface. The time response curve refers to the function curve of the dynamic diffraction coefficient changing with time. The delay time refers to the time interval from the initial moment the pulse generator emits a pulse to the moment when the pulse interacts with the defect location and produces a measurable effect.

[0069] Optionally, the pulse wavefront phase distribution of the pulse generator can be determined by a pulse propagation mathematical model, such as the Huygens-Fresnel principle, diffraction integral, etc.

[0070] Optionally, the time response curve of the dynamic diffraction coefficient may be constructed by a nonlinear least squares method, such as a Levenberg-Marquardt algorithm.

[0071] S5. Combining the pulse wavefront phase distribution and the trigger timing, the decision module of the AI ​​adsorption defect analysis model is used to determine the pulse optimization scheme of the component to be packaged and the packaging film during the packaging process, so as to perform the packaging of the component to be packaged and obtain the target packaged component.

[0072] By combining the pulse wavefront phase distribution and the trigger timing, the embodiments of the present invention, through the decision module of the AI ​​adsorption defect analysis model, determine a pulse optimization scheme for the component to be packaged and the encapsulation film during the encapsulation process. This ensures more uniform and precise deposition of laser energy in the contact area between the component to be packaged and the encapsulation film, better adapts to variations in adsorption force, reduces adsorption defects, and thus improves yield consistency between and within batches. The pulse optimization scheme refers to the specific instructions output by the decision module of the AI ​​adsorption defect analysis model.

[0073] By performing the packaging of the target component, the present invention can more effectively address defective areas of the target packaged component, achieving more controlled repair and enhancing the accuracy and controllability of the packaging process. The target packaged component refers to the packaged component that is successfully manufactured by executing the pulse optimization scheme described above.

[0074] Compared with the problems described in the background technology, the embodiment of the present invention can achieve true process visibility and transparency by real-time collection of multi-source data between the component to be packaged and the packaging film, thereby improving the timeliness and accuracy of defect detection; optionally, the embodiment of the present invention can fuse the aligned multi-source data to obtain a comprehensive data set, which can complement each other and make up for each other's shortcomings, thereby obtaining a more accurate and robust description than a single sensor, covering a wider space, and filling in the regional information that a single sensor cannot observe; the embodiment of the present invention constructs an adsorption force evolution network between the component to be packaged and the packaging film according to the adsorption force and the adsorption force type through the network evolution module of the AI ​​adsorption defect analysis model, which can simulate the dynamic process of the adsorption force size, distribution and different types of adsorption force changing with environmental factors such as time, temperature, pressure, humidity, etc. from the beginning of contact to the completion of packaging and even the subsequent possible use process, so as to analyze the evolution of the adsorption force under specific process parameters or environmental conditions, as well as the critical point of this evolution, thereby providing Early warning of potential defects; by calculating the probability of adsorption defects between the component to be packaged and the packaging film, embodiments of the present invention can achieve accurate defect prediction and risk quantification, thereby significantly reducing defective products caused by adsorption defects in the final product; by determining the pulse wavefront phase distribution and trigger timing of the pulse generator, embodiments of the present invention can ensure that energy is more concentrated on the target area (e.g., the defect point), improving processing efficiency and effectiveness, reducing damage to surrounding areas, and precisely controlling the processes of energy accumulation, thermal effects, and chemical reactions, thereby reducing random fluctuations in pulse characteristics; finally, by combining the pulse wavefront phase distribution and trigger timing, embodiments of the present invention determine the pulse optimization scheme for the component to be packaged and the packaging film during the packaging process through the decision module of the AI ​​adsorption defect analysis model, ensuring more uniform and precise deposition of laser energy in the contact area between the component to be packaged and the packaging film, better adapting to variations in adsorption force, reducing adsorption defects, and thus improving yield consistency between and within batches. Therefore, the AI-based method for reducing packaging adsorption defects provided by embodiments of the present invention can improve product quality and efficiency during the packaging process.

[0075] Example 2: like Figure 2 The figure shows a functional module diagram of a system for reducing packaging adsorption defects based on AI intelligence according to the present invention.

[0076] The AI-based packaging adsorption defect reduction system 200 described in the present invention can be installed in an electronic device. Depending on the functionality implemented, the AI-based packaging adsorption defect reduction system can include a multi-source data acquisition module 201, a multi-source data fusion module 202, an adsorption defect probability calculation module 203, a pulse parameter analysis module 204, and a packaging optimization module 205. The modules described in the present invention, also referred to as units, are a series of computer program segments that can be executed by an electronic device processor and perform fixed functions, and are stored in the electronic device's memory.

[0077] In the embodiment of the present invention, the functions of each module / unit are as follows: The multi-source data acquisition module 201 is used to collect multi-source data between the components to be packaged and the packaging film in real time; The multi-source data fusion module 202 is used to perform spatiotemporal alignment on the multi-source data to obtain aligned multi-source data, and perform data fusion on the aligned multi-source data to obtain a comprehensive data set; The adsorption defect probability calculation module 203 is configured to analyze the adsorption force between the component to be packaged and the packaging film based on the comprehensive data set using a defect analysis module of a preset AI adsorption defect analysis model, determine the adsorption force type of the adsorption force, and construct an adsorption force evolution network between the component to be packaged and the packaging film based on the adsorption force and the adsorption force type using a network evolution module of the AI ​​adsorption defect analysis model to calculate the adsorption defect probability between the component to be packaged and the packaging film; The pulse parameter analysis module 204 is configured to analyze the influencing factors of the adsorption defect probability, configure a pulse generator for the component to be packaged and the packaging film during the packaging process, and calculate the pulse parameters of the pulse generator based on the influencing factors, wherein the pulse parameters include: injection position, pulse pressure, and pulse duration. The dynamic diffraction coefficient of the pulse generator is calculated by combining the pulse parameters and the comprehensive data set to determine the pulse wavefront phase distribution and trigger timing of the pulse generator. The packaging optimization module 205 is used to combine the pulse wavefront phase distribution and the trigger timing, and determine the pulse optimization scheme of the component to be packaged and the packaging film during the packaging process through the decision module of the AI ​​adsorption defect analysis model, so as to perform the packaging of the component to be packaged and obtain the target packaged component.

[0078] In detail, the modules in the AI-based system for reducing package adsorption defects 200 in the embodiment of the present invention are used in the same manner as above. Figure 1The same technical means as the method of reducing packaging adsorption defects based on AI intelligence described in the article can produce the same technical effects, so I will not go into details here.

[0079] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0080] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for reducing packaging adsorption defects based on AI intelligence, characterized in that: The method comprises: Real-time collection of multi-source data between the components to be packaged and the packaging film; Performing spatiotemporal alignment on the multi-source data to obtain aligned multi-source data, and performing data fusion on the aligned multi-source data to obtain a comprehensive data set; analyzing, based on the comprehensive data set, the adsorption force between the component to be packaged and the packaging film using a defect analysis module of a preset AI adsorption defect analysis model to determine an adsorption force type of the adsorption force; and constructing, based on the adsorption force and the adsorption force type, an adsorption force evolution network between the component to be packaged and the packaging film using a network evolution module of the AI ​​adsorption defect analysis model to calculate a probability of adsorption defects between the component to be packaged and the packaging film; Analyzing the factors influencing the adsorption defect probability, configuring a pulse generator for the component to be packaged and the packaging film during the packaging process, and calculating pulse parameters of the pulse generator based on the influencing factors, wherein the pulse parameters include: injection position, pulse pressure, and pulse duration. Combining the pulse parameters with the comprehensive data set, calculating the dynamic diffraction coefficient of the pulse generator to determine the pulse wavefront phase distribution and trigger timing of the pulse generator; In combination with the pulse wavefront phase distribution and the trigger timing, the decision module of the AI ​​adsorption defect analysis model is used to determine the pulse optimization scheme of the component to be packaged and the packaging film during the packaging process to perform the packaging of the component to be packaged and obtain the target packaged component.

2. The method for reducing package adsorption defects based on AI intelligence according to claim 1, characterized in that: The real-time collection of multi-source data between the component to be packaged and the packaging film includes: Clarify the data collection target required between the component to be packaged and the packaging film; Based on the data acquisition target, configuring multiple sensors between the component to be packaged and the packaging film; Determining a connection method and network layout of the multiple sensors; constructing a sensor network of the multiple sensors according to the connection mode and the network layout; Determining a sampling rate of the multi-sensor to construct a time synchronization network of the multi-sensor; Multi-source data between the component to be packaged and the packaging film is collected according to the sensor network and the time synchronization network.

3. The method for reducing package adsorption defects based on AI intelligence according to claim 1, characterized in that: The performing spatiotemporal alignment on the multi-source data to obtain aligned multi-source data includes: Preprocessing the multi-source data to obtain preprocessed multi-source data; Determining a main time axis of the pre-processed multi-source data and identifying a timestamp of the pre-processed multi-source data; performing time alignment on the preprocessed multi-source data according to the main time axis and the timestamp to obtain time-aligned multi-source data; Constructing a global reference coordinate system for the time-aligned multi-source data and extracting local coordinate information of the time-aligned multi-source data; Converting the local coordinate information into the global reference coordinate system to obtain global coordinates; Based on the global coordinates, spatial registration is performed on the time-aligned multi-source data to obtain aligned multi-source data.

4. The method for reducing package adsorption defects based on AI intelligence according to claim 1, characterized in that: Analyzing the adsorption force between the component to be packaged and the packaging film using a defect analysis module of a preset AI adsorption defect analysis model based on the comprehensive data set includes: Extracting key characteristic parameters of the comprehensive data set, wherein the key characteristic parameters include: material properties, process parameters, surface parameters, and environmental parameters; Based on the key characteristic parameters, the adsorption force between the component to be encapsulated and the encapsulation film is calculated using a hybrid adsorption force algorithm in the defect analysis module, wherein the hybrid adsorption force algorithm includes: ； in, Indicates adsorption force, Indicates the strength of molecular interaction between materials in material properties, Indicates the surface distance correction factor between the component to be packaged and the packaging film, represents pi, Indicates the average gap distance between the component to be packaged and the packaging film, represents the equivalent contact area correction factor, Indicates the actual contact area radius in the process parameters, represents the cosine function, Represents the contact angle in the process parameters, represents the surface energy among the surface parameters, Indicates The exponential function with base , represents the fitting parameters, Indicates the relative humidity in the environmental parameters. represents the elastic deformation coefficient, represents the Young's modulus of the film, Indicates the peak-to-valley height of the surface.

5. The method for reducing package adsorption defects based on AI intelligence according to claim 1, characterized in that: The step of constructing an adsorption force evolution network between the component to be packaged and the packaging film according to the adsorption force and the adsorption force type by using a network evolution module of the AI ​​adsorption defect analysis model includes: dividing the network nodes between the components to be packaged and the packaging film; constructing a node feature vector of the network node according to the adsorption force and the adsorption force type; defining edge connection rules for the network nodes; Determining the adsorption force transmission strength between the network nodes according to the edge connection rule; Based on the node feature vector and the adsorption force transmission strength, a time evolution algorithm of the network node is defined, wherein the time evolution algorithm includes: ； in, Represents a network node In time The node feature vector of represents the activation function, represents the physical constraint matrix, Represents a network node In time The node feature vector of Represents a network node The neighbor set of represents the bias term, represents the relaxation time, Represents a network node The node feature vector of The instantaneous rate of change of Determining the instantaneous rate of change of the node feature vector according to the time evolution algorithm; According to the instantaneous change rate, an adsorption force evolution network between the component to be packaged and the packaging film is constructed.

6. The method for reducing package adsorption defects based on AI intelligence according to claim 1, characterized in that: The calculating of the probability of adsorption defects between the component to be packaged and the packaging film includes: Analyzing the change rate of the node adsorption force in the corresponding adsorption force evolution network between the component to be encapsulated and the encapsulation film; Determining attention weights between corresponding network nodes of the adsorption force evolution network; Calculating a node adsorption defect probability of the network node according to the attention weight and the node adsorption force change rate; Identifying a boundary of the component to be packaged, and calculating a normalized distance between the network node and the boundary; The adsorption defect probability between the component to be packaged and the packaging film is calculated according to the normalized distance and the node adsorption defect probability.

7. The method for reducing package adsorption defects based on AI intelligence according to claim 1, characterized in that: The calculating the pulse parameters of the pulse generator based on the influencing factors includes: Fitting a pulse parameter analysis model of the pulse generator based on the influencing factors; Determining the constraints and objective function of the pulse parameter analysis model; constructing an iterative mechanism for the pulse parameter analysis model according to the constraint conditions and the objective function; determining initial pulse parameters of the pulse parameter analysis model, and analyzing convergence of the initial pulse parameters; When the convergence does not meet a preset convergence threshold, optimizing the initial pulse parameters through the iterative mechanism to obtain optimized pulse parameters; Analyzing the post-optimization convergence of the optimized pulse parameters; When the convergence after optimization meets the convergence threshold, the optimized pulse parameters are used as the pulse parameters of the pulse generator.

8. The method for reducing package adsorption defects based on AI intelligence according to claim 1, characterized in that: The step of calculating the dynamic diffraction coefficient of the pulse generator by combining the pulse parameters and the comprehensive data set comprises: extracting comprehensive data features of the comprehensive data set; Performing feature normalization fusion on the comprehensive data features and the pulse parameters to obtain normalized fusion features; Calculating the fluctuation field value of the pulse generator according to the normalized fusion feature; Calculating the defect area energy and the total incident energy of the pulse generator according to the fluctuation field value; The dynamic diffraction coefficient of the pulse generator is calculated based on the defect area energy and the incident total energy.

9. The method for reducing package adsorption defects based on AI intelligence according to claim 8, characterized in that: Determining the pulse wavefront phase distribution and trigger timing of the pulse generator includes: determining a pulse wavelength of the pulse generator; Identifying a defect position corresponding to the pulse generator and analyzing a defect height difference at the defect position; Determining the pulse wavefront phase distribution of the pulse generator according to the pulse wavelength, the defect height difference and the dynamic diffraction coefficient corresponding to the pulse generator; constructing a time response curve of the dynamic diffraction coefficient; Calculating the delay time of the pulse corresponding to the pulse generator based on the time response curve; The triggering timing of the pulse generator is determined according to the delay time.

10. The AI-based system for reducing package adsorption defects according to claim 1, wherein: The system comprises: Multi-source data acquisition module, used to collect multi-source data between the components to be packaged and the packaging film in real time; A multi-source data fusion module is used to perform spatiotemporal alignment on the multi-source data to obtain aligned multi-source data, and perform data fusion on the aligned multi-source data to obtain a comprehensive data set; an adsorption defect probability calculation module, configured to analyze, based on the comprehensive data set and using a defect analysis module of a preset AI adsorption defect analysis model, the adsorption force between the component to be packaged and the packaging film, determine the adsorption force type of the adsorption force, and construct, based on the adsorption force and the adsorption force type, an adsorption force evolution network between the component to be packaged and the packaging film using a network evolution module of the AI ​​adsorption defect analysis model to calculate the adsorption defect probability between the component to be packaged and the packaging film; a pulse parameter analysis module, configured to analyze factors influencing the adsorption defect probability, configure a pulse generator for the component to be packaged and the packaging film during the packaging process, and calculate pulse parameters of the pulse generator based on the influencing factors, wherein the pulse parameters include: injection position, pulse pressure, and pulse duration; and calculate the dynamic diffraction coefficient of the pulse generator by combining the pulse parameters and the comprehensive data set to determine the pulse wavefront phase distribution and trigger timing of the pulse generator; A packaging optimization module is used to combine the pulse wavefront phase distribution and the trigger timing, and determine the pulse optimization scheme of the component to be packaged and the packaging film during the packaging process through the decision module of the AI ​​adsorption defect analysis model, so as to perform the packaging of the component to be packaged and obtain the target packaged component.

Images