Method and system for detecting packaging defect of industrial solid state disk

Abstract

The invention relates to the technical field of high-end chips, and discloses a packaging defect detection method and system for an industrial solid state disk, and the method comprises the steps: collecting multi-source heterogeneous detection data, and extracting the multi-dimensional defect features of a packaging body; packaging process parameters and material attributes of the industrial solid state disk are input into the virtual mirror image body, and a defect sensitivity distribution diagram of the packaging body is generated; identifying a defect expansion path and a defect type of the packaging body; based on the process weak link, the defect expansion path and the defect type, generating a grading defect detection instruction of the packaging body; setting a personalized defect detection mode of the packaging body in combination with the number of NAND stacked layers, the solder ball spacing and the defect type; and outputting a defect detection result of the packaging body through the virtual mirror image body by utilizing a grading defect detection instruction and a personalized defect detection mode. According to the method, the accuracy and credibility of an industrial solid state disk packaging defect detection result can be effectively improved.

Description

Technical Field

[0001] The present invention relates to a packaging defect detection method and system for an industrial solid-state hard disk, belonging to the technical field of high-end chips. Background Art

[0002] An industrial solid-state drive (SSD) is a storage device designed specifically for extreme environments. It utilizes a special master control chip, high-quality flash memory particles, and a reinforced design, combined with high-speed data transmission protocols and advanced error correction algorithms to enable fast data reading and writing and reliable storage. Due to its resistance to high and low temperatures, vibration and shock, and low power consumption, it is widely used in harsh environments such as industrial automation control, vehicle-mounted monitoring, and aerospace. With the increasing requirements for read and write speeds, shock resistance, and service life of data storage devices in industrial automation, edge computing, and other fields, the reliability of industrial SSDs in extreme environments is becoming increasingly critical. Therefore, in order to meet the demands of industrial-grade applications for highly stable and durable storage devices, it is extremely important to build a high-precision and efficient packaging defect detection system.

[0003] Existing methods for detecting defects in industrial SSD packaging mainly rely on automated optical inspection methods, which automatically scan for appearance defects in industrial SSDs using high-resolution cameras and image algorithms. While this method can improve the speed of defect detection, it cannot monitor defect expansion in real time (such as crack extension under thermal cycling stress), potentially causing early defects to evolve into functional failures, leading to a decline in the overall quality of industrial SSDs.

[0004] Therefore, there is an urgent need for a solution that can effectively improve the accuracy and reliability of industrial solid-state drive packaging defect detection results. Summary of the Invention

[0005] The present invention provides a method and system for detecting packaging defects of industrial solid-state hard drives, the main purpose of which is to effectively improve the accuracy and credibility of industrial solid-state hard drive packaging defect detection results.

[0006] To achieve the above objectives, the present invention provides a method for detecting packaging defects of industrial solid-state drives, comprising: Obtaining a package of an industrial solid-state drive, collecting multi-source heterogeneous detection data of the package, and extracting multi-dimensional defect features of the package based on the multi-source heterogeneous detection data; Constructing a virtual mirror image of the package using the multi-source heterogeneous detection data, and inputting the packaging process parameters and material properties of the industrial solid-state drive into the virtual mirror image to generate a defect sensitivity distribution map of the package; collecting multi-parameter time series data of the package in real time, identifying a defect extension path of the package based on the multi-parameter time series data and the virtual mirror, and determining a defect type of the package according to the multi-dimensional defect characteristics; Locating process weaknesses of the package according to the defect type and the defect sensitivity distribution map, and generating a graded defect detection instruction for the package based on the process weaknesses, the defect extension path, and the defect type; querying the hardware configuration of the industrial solid-state drive, extracting the number of NAND stacking layers and the solder ball pitch in the hardware configuration, and setting a personalized defect detection method for the package body based on the number of NAND stacking layers, the solder ball pitch, and the defect type; In combination with the hierarchical defect detection instruction and the personalized defect detection method, the defect detection result of the package body is output through the virtual mirror body.

[0007] Optionally, extracting multi-dimensional defect features of the package based on the multi-source heterogeneous detection data includes: Performing spatiotemporal registration processing on the multi-source heterogeneous detection data to obtain target fusion data; extracting image data, electrical parameter data, acoustic image data, and environmental response data from the target fusion data; performing image enhancement processing on the image data to obtain enhanced image data; Segmenting a defect area from the enhanced image data and extracting defect geometric features of the defect area; Performing outlier elimination and smoothing processing on the electrical parameter data to obtain a target electrical parameter sequence; Analyzing parameter change trends in the target electrical parameter sequence to extract abnormal electrical fluctuation characteristics in the target electrical parameter sequence; Performing variational modal decomposition processing on the ultrasonic signal in the acoustic image data to obtain a characteristic frequency band, and calculating a frequency band energy ratio of the characteristic frequency band; extracting environmental sensitivity characteristics of the package based on the environmental response data; The multi-dimensional defect characteristics of the package are determined based on the defect geometric characteristics, the abnormal electrical fluctuation characteristics, the frequency band energy proportion and the environmental sensitivity characteristics.

[0008] Optionally, the constructing a virtual mirror image of the package body using the multi-source heterogeneous detection data includes: Determining structural features of the package and corresponding material states thereof through the multi-source heterogeneous detection data, and identifying environmental load data of the package; Analyzing the dynamic response relationship between the structural characteristics and the environmental load data; identifying failure modes of the package under different working conditions according to the dynamic response relationship; Based on the failure mode, setting a damage evolution network of the package under the different working conditions; constructing a parameterized geometric structure of the package according to the structural features and the material state; Performing multi-physics field attribute assignment processing on the parameterized geometric architecture based on the physical characteristics of the multi-source heterogeneous detection data to obtain multi-physics field attributes; Combining the damage evolution network, the parameterized geometric framework and the multi-physics field properties, a virtual mirror body of the package is constructed.

[0009] Optionally, inputting the packaging process parameters and material properties of the industrial solid-state drive into the virtual mirror body to generate a defect sensitivity distribution map of the package body includes: After inputting the packaging process parameters and material properties of the industrial solid-state drive into the virtual mirror body, identifying the energy distribution of the virtual mirror body; Calculating defect sensitivity indexes of different regions of the virtual mirror body based on the energy distribution; According to the defect sensitivity index, setting the defect sensitivity level of the different areas; Based on the defect sensitivity level, defining visualization color mapping rules for the different areas; Visualization processing of the different regions is performed according to the visualization color mapping rule to obtain a defect sensitivity distribution map of the package.

[0010] Optionally, identifying the defect extension path of the package based on the multi-parameter timing data and the virtual mirror includes: Loading the multi-parameter time series data through the virtual mirror body, performing multi-physics field coupling simulation processing on the package body, and obtaining a stress distribution simulation result; outputting key defect driving parameters of the package based on the stress distribution simulation results; Analyzing the defect expansion trend of the package body driven by the key defect driving parameters; defining an adaptive defect expansion threshold for the package; marking a critical defect expansion area of ​​the package according to the adaptive defect expansion threshold and the defect expansion trend; Extracting multi-parameter time series data features of the key defect area to construct a defect extension feature vector of the key defect area; Obtaining a defect extension path segment of the package based on the defect extension feature vector; The defect extension path segment is fitted by a path fitting algorithm to obtain a defect extension path.

[0011] Optionally, determining the defect type of the package according to the multi-dimensional defect characteristics includes: Identifying macroscopic structural features, mesoscopic microstructural unit features, and microstructural features of the package from the multidimensional defect features; performing cross-scale correlation processing of the macroscopic structural features, the mesoscopic microstructure unit features, and the microstructure features to obtain a fused defect feature list; Calculating the feature importance coefficient and cross-scale defect correlation of each feature in the fused defect feature list; Extracting core defect features of the package based on the feature importance coefficient and the cross-scale defect correlation; The defect type of the package is determined based on the core defect characteristics.

[0012] Optionally, locating a process weak link of the package body according to the defect type and the defect sensitivity distribution map includes: Calculate the occurrence frequency of the defect type, and filter out high-frequency defect categories based on the occurrence frequency; Identifying a defect location corresponding to the high-frequency defect category and obtaining point cloud data of the defect location; Performing spatial registration processing on the point cloud data and the defect sensitivity distribution map to obtain a spatial registration result; Based on the spatial registration result, identifying an overlapping area between the point cloud data and the defect sensitivity distribution map; Retrieving historical process parameter data corresponding to the package according to the overlapping area; Analyzing the correlation between the process parameter historical data and the defect type; Based on the association relationship, identifying key process steps and operating parameters corresponding to the defect type; The process weaknesses of the package are located through the key process steps and operating parameters.

[0013] Optionally, the generating of a graded defect detection instruction for the package body based on the process weak link, the defect extension path, and the defect type includes: Identify the abnormal frequency of process parameters corresponding to the weak links in the process; Calculating the probability of defect occurrence in each area of ​​the package body according to the abnormal frequency of the process parameters; identifying a defect diffusion rate of the package body in the defect propagation path; Determining the degree of damage caused by the defect of the package according to the defect type; Establishing a three-dimensional risk matrix for the package body according to the defect occurrence probability, the defect diffusion speed, and the defect hazard degree; determining a defect risk level of the package based on the three-dimensional risk matrix; Develop an executable detection protocol corresponding to the defect risk level; A graded defect inspection instruction for the package is generated according to the defect risk level and the executable inspection protocol.

[0014] Optionally, the step of setting a personalized defect detection method for the package body in combination with the number of NAND stacking layers, the solder ball pitch, and the defect type includes: Extracting the interlayer interconnection density and thermal stress distribution characteristics of the package according to the number of NAND stack layers; Calculating the micro solder joint detection accuracy of the package body according to the solder ball spacing; Identifying a key inspection area of ​​the package body based on the interlayer interconnection density, the thermal stress distribution characteristics, and the micro-solder point inspection accuracy; Based on the defect type, analyzing the defect formation mechanism and detection sensitivity requirements of the package; Establishing a defect feature database of the package using the defect formation mechanism and the detection sensitivity requirement; Creating a multimodal collaborative inspection scheme for the package according to the key inspection area and the defect feature database; Identify the detection modalities corresponding to the multimodal collaborative detection solution and define the timing coordination rules between the detection modalities; According to the multimodal collaborative detection scheme and the timing coordination rule, a personalized defect detection method for the package body is set.

[0015] In order to solve the above problems, the present invention also provides a packaging defect detection system for industrial solid-state hard drives, the system comprising: a data acquisition module, configured to acquire a package of an industrial solid-state drive, collect multi-source heterogeneous detection data of the package, and extract multi-dimensional defect features of the package based on the multi-source heterogeneous detection data; a virtual mirror body construction module, configured to construct a virtual mirror body of the package body using the multi-source heterogeneous detection data, and input the packaging process parameters and material properties of the industrial solid-state drive into the virtual mirror body to generate a defect sensitivity distribution map of the package body; a defect analysis module configured to collect multi-parameter time series data of the package in real time, identify a defect extension path of the package based on the multi-parameter time series data and the virtual mirror, and determine a defect type of the package based on the multi-dimensional defect characteristics; a detection instruction generation module, configured to locate a process weakness of the package according to the defect type and the defect sensitivity distribution map, and generate a graded defect detection instruction for the package based on the process weakness, the defect extension path, and the defect type; an execution mode determination module, configured to query the hardware configuration of the industrial solid-state drive, extract the number of NAND stacking layers and the solder ball pitch in the hardware configuration, and set a personalized defect detection mode for the package body based on the number of NAND stacking layers, the solder ball pitch, and the defect type; A result output module is configured to combine the graded defect detection instruction and the personalized defect detection method to output the defect detection result of the package through the virtual mirror body.

[0016] Compared with the problems described in the background technology, the embodiment of the present invention collects multi-source heterogeneous detection data of the package body and extracts multi-dimensional defect characteristics of the package body based on the multi-source heterogeneous detection data, which can provide data support for the optimization of the industrial solid-state hard disk packaging process; further, the embodiment of the present invention constructs a virtual mirror body of the package body by utilizing the multi-source heterogeneous detection data, which can accurately locate the weak links in the process and formulate personalized detection strategies, thereby effectively improving the pertinence and accuracy of the defect detection of the industrial solid-state hard disk package body. The embodiment of the present invention inputs the packaging process parameters and material properties of the industrial solid-state hard disk into the virtual mirror body The invention can generate a defect sensitivity distribution map of the package body, and can intuitively find high-incidence areas of defects such as cold solder joints and chip delamination in the package body, quickly determine the detection focus, reduce unnecessary detection steps, and improve the defect detection efficiency of the industrial solid-state hard disk package body; the embodiment of the present invention can identify the defect expansion path of the package body based on the multi-parameter time series data and the virtual mirror body, and can dynamically associate the causal relationship between environmental load and defect evolution, optimize the design and material selection of the industrial solid-state hard disk package body, and further, the embodiment of the present invention can determine the defect type of the package body according to the multi-dimensional defect characteristics, and can identify the defect type of the industrial solid-state hard disk package body. The defects are converted from fuzzy judgment to precise classification, thereby improving the pertinence of defect detection of industrial solid-state hard disk packages; the embodiment of the present invention locates the process weaknesses of the package according to the defect type and the defect sensitivity distribution map, and can quickly focus on the key risk points of the industrial solid-state hard disk packaging process to optimize the corresponding process parameter settings, thereby effectively reducing the defect rate in the industrial solid-state hard disk packaging process and improving the overall product quality and production efficiency. Furthermore, the embodiment of the present invention generates a hierarchical defect detection instruction for the package by combining the process weaknesses, the defect extension path and the defect type, which can effectively integrate detection resources. Avoid efficiency loss caused by blind detection; the embodiment of the present invention queries the hardware configuration of the industrial solid-state drive and extracts the number of NAND stacking layers and solder ball pitch in the hardware configuration. It can adjust the defect detection strategy in a targeted manner according to the characteristics of different numbers of stacking layers and solder ball pitches, significantly improving the detection efficiency and the accuracy of defect identification. Furthermore, the embodiment of the present invention sets a personalized defect detection method for the package body by combining the number of NAND stacking layers, the solder ball pitch and the defect type. It can achieve deep adaptation of the detection strategy with hardware characteristics and defect characteristics, accurately locate process risk points, and significantly improve the efficiency of package defect detection.Finally, the embodiment of the present invention combines the hierarchical defect detection instructions with the personalized defect detection method, and outputs the defect detection results of the package through the virtual mirror. This not only enables the precise allocation of detection resources according to the defect risk level and package characteristics, significantly improving the defect detection rate in high-risk areas, but also enables the use of the virtual mirror to simulate and verify the detection process and results in real time, predict detection blind spots and potential errors in advance, achieve dynamic optimization of detection strategies, and effectively improve the accuracy and reliability of detection results. Therefore, the embodiment of the present invention provides a method and system for industrial solid-state drive package defect detection, which can effectively improve the accuracy and reliability of industrial solid-state drive package defect detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A schematic flow chart of a method for detecting packaging defects of industrial solid-state drives provided by one embodiment of the present invention; Figure 2 A schematic diagram of a NAND stack structure for a method for detecting packaging defects of an industrial solid-state drive provided by one embodiment of the present invention; Figure 3 This is a functional module diagram of a packaging defect detection system for an industrial solid-state drive provided in one embodiment of the present invention.

[0018] 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

[0019] 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.

[0020] An embodiment of the present application provides a method for detecting packaging defects in industrial solid-state drives. The execution subject of the method 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 in the embodiment of the present application. In other words, the method for detecting packaging defects in industrial solid-state drives 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.

[0021] Example 1 Reference Figure 1 FIG. 1 is a flow chart of a method for detecting packaging defects of an industrial solid-state drive provided by an embodiment of the present invention. In this embodiment, the method for detecting packaging defects of an industrial solid-state drive includes: S1. Obtain a package of an industrial solid-state drive, collect multi-source heterogeneous detection data of the package, and extract multi-dimensional defect features of the package based on the multi-source heterogeneous detection data.

[0022] By obtaining the package of an industrial solid-state drive, the embodiment of the present invention can accurately locate the type and root cause of packaging defects, thereby more efficiently identifying internal defects in the packaging structure. The industrial solid-state drive refers to a storage device designed specifically for harsh industrial environments. It mainly uses flash memory chips as storage media and performs data read and write control and management through a main control chip. The package refers to an overall structural unit that uses shell materials such as metal, ceramic or engineering plastics to seal and protect the core electronic components of the industrial solid-state drive and the accompanying printed circuit board.

[0023] Furthermore, the embodiments of the present invention can provide data support for the optimization of industrial solid-state drive packaging processes by collecting multi-source heterogeneous detection data of the package and extracting multi-dimensional defect features of the package based on the multi-source heterogeneous detection data. The multi-source heterogeneous detection data refers to a data set with different formats, structures and properties obtained through a variety of different detection technologies, equipment and channels, wherein the detection technologies include X-ray tomography (CT), ultrasonic guided wave detection and terahertz imaging, etc. The multi-dimensional defect features refer to feature information extracted from the multi-source heterogeneous detection data that can comprehensively and meticulously describe the defects of the industrial solid-state drive package in multiple dimensions, including solder joint void size, substrate delamination area and material interface defects.

[0024] As an embodiment of the present invention, extracting multi-dimensional defect features of the package based on the multi-source heterogeneous inspection data includes: Performing spatiotemporal registration processing on the multi-source heterogeneous detection data to obtain target fusion data; extracting image data, electrical parameter data, acoustic image data, and environmental response data from the target fusion data; performing image enhancement processing on the image data to obtain enhanced image data; Segmenting a defect area from the enhanced image data and extracting defect geometric features of the defect area; Performing outlier elimination and smoothing processing on the electrical parameter data to obtain a target electrical parameter sequence; Analyzing parameter change trends in the target electrical parameter sequence to extract abnormal electrical fluctuation characteristics in the target electrical parameter sequence; Performing variational modal decomposition processing on the ultrasonic signal in the acoustic image data to obtain a characteristic frequency band, and calculating a frequency band energy ratio of the characteristic frequency band; extracting environmental sensitivity characteristics of the package based on the environmental response data; The multi-dimensional defect characteristics of the package are determined based on the defect geometric characteristics, the abnormal electrical fluctuation characteristics, the frequency band energy proportion and the environmental sensitivity characteristics.

[0025] Among them, the target fusion data refers to the integrated, aligned and optimized result data obtained after performing spatiotemporal registration processing on multi-source heterogeneous detection data, wherein the spatiotemporal registration processing of the target fusion data includes spatial registration processing and spatiotemporal synchronization processing, wherein the spatial registration processing can be implemented by the ICP algorithm, and the spatial registration processing can be implemented by the PTPv2 protocol, the image data refers to the image information of the industrial solid-state hard disk package obtained by optical imaging, X-ray imaging and other technologies, including the appearance and internal structure of the package, etc., the electrical parameter data refers to the data related to electrical performance collected when the electrical performance test of the industrial solid-state hard disk package is performed, such as voltage, current, resistance, capacitance, impedance, signal transmission delay, etc., the acoustic image data refers to the image information of the industrial solid-state hard disk package obtained by using ultrasonic scanning microscope and other equipment The device transmits ultrasonic waves to the industrial solid-state drive package, receives the reflected ultrasonic signals and converts them into image data. The environmental response data refers to the change data of the performance parameters of the industrial solid-state drive package collected under simulated environmental conditions (such as high and low temperatures, humidity changes, vibration, electromagnetic interference, etc.). The enhanced image data refers to the image data obtained after optimizing the original image data using an image enhancement algorithm (such as histogram equalization, filtering, contrast stretching, etc.). The defect area refers to the partial area where defects exist in the enhanced image data, wherein the defect area can be segmented using a U-Net3+ network. The defect geometric features refer to the feature parameters that describe the geometric properties of the defect, such as the shape, size, and position of the defect. For example, the void defect features may include the shape (circular, elliptical, etc.), diameter, and volume of the void.The characteristics of delamination defects include the area, position, and direction of the delamination interface. The target electrical parameter sequence refers to a stable and reliable electrical parameter sequence obtained by removing outliers and smoothing the original electrical parameter data based on the 3σ criterion and the DBSCAN clustering method. The 3σ criterion is used to identify and remove abnormal data points that deviate significantly from the normal range, and the DBSCAN clustering algorithm can further process the noise and discrete points in the data, so that the electrical parameter sequence can more accurately reflect the true electrical performance of the package. The parameter change trend refers to the change law and trend of each parameter in the target electrical parameter sequence over time or other variables, such as the trend of the resistance value gradually increasing or decreasing with the test time. The abnormal electrical fluctuation characteristics refer to the frequency, amplitude, duration and other characteristics of abnormal fluctuations in the parameters in the target electrical parameter sequence, such as a sudden and large fluctuation in the current value. The ultrasonic signal refers to the ultrasonic scanning equipment sending a large fluctuation to the industrial solid-state hard drive package. The package transmits and receives signals carried by reflected ultrasonic waves. Variational mode decomposition (VMD) is an adaptive, non-recursive signal processing method used to decompose complex signals into multiple eigenmode functions with specific center frequencies and bandwidths. The characteristic frequency band refers to the frequency range corresponding to specific types of defects (such as substrate delamination and solder voids) in the frequency domain data of the ultrasonic signal after variational mode decomposition. For example, the frequency range corresponding to substrate delamination can be 2.3-2.7 MHz, and the frequency range corresponding to solder voids can be 3.5-4.2 MHz. The frequency band energy fraction refers to the proportion of the signal energy within the characteristic frequency band to the total energy of the ultrasonic signal. The environmental sensitivity characteristic refers to a characteristic parameter based on environmental response data that reflects the sensitivity of the industrial solid-state drive package to performance changes under different environmental conditions. For example, the decrease in the package's read and write speed in a high-temperature environment or the fluctuation of electrical parameters when humidity changes.

[0026] S2. Utilize the multi-source heterogeneous detection data to construct a virtual mirror image of the package, and input the packaging process parameters and material properties of the industrial solid-state drive into the virtual mirror image to generate a defect sensitivity distribution map of the package.

[0027] By utilizing the multi-source heterogeneous detection data to construct a virtual mirror image of the package, the embodiment of the present invention can accurately locate process weaknesses and formulate personalized detection strategies, thereby effectively improving the pertinence and accuracy of defect detection of industrial solid-state drive packages. The virtual mirror image refers to a digital model that is highly corresponding to the physical package and is constructed in a virtual space based on the multi-source heterogeneous detection data of the industrial solid-state drive package through digital twin technology.

[0028] As an embodiment of the present invention, the step of constructing a virtual image of the package using the multi-source heterogeneous detection data includes: Determining structural features of the package and corresponding material states thereof through the multi-source heterogeneous detection data, and identifying environmental load data of the package; Analyzing the dynamic response relationship between the structural characteristics and the environmental load data; identifying failure modes of the package under different working conditions according to the dynamic response relationship; Based on the failure mode, setting a damage evolution network of the package under the different working conditions; constructing a parameterized geometric structure of the package according to the structural features and the material state; Performing multi-physics field attribute assignment processing on the parameterized geometric architecture based on the physical characteristics of the multi-source heterogeneous detection data to obtain multi-physics field attributes; Combining the damage evolution network, the parameterized geometric framework and the multi-physics field properties, a virtual mirror body of the package is constructed.

[0029] The structural features refer to the physical structure-related characteristics of the industrial solid-state drive package, including the overall dimensions of the package, internal chip layout, solder joint distribution and shape, substrate layer number and structure, package shell type and construction, etc. The material state refers to the properties and conditions of the materials of each component in the package, including the type of material (such as the semiconductor material of the flash memory chip, the epoxy resin material of the package substrate, the solder alloy material of the solder joint, etc.), physical performance parameters (such as density, thermal expansion coefficient, elastic modulus, thermal conductivity, electrical conductivity, etc.), chemical properties (such as oxidation resistance, corrosion resistance), and possible state changes of the material during manufacturing and use (such as aging, fatigue, and damage degree). The environmental load data refers to the data on various environmental factors that the industrial solid-state drive package may be subjected to during actual use, including temperature variation range (such as extreme high temperature and low temperature conditions), humidity fluctuations, frequency and intensity of vibration and shock, intensity and frequency band of electromagnetic interference, air pressure changes, etc. Optionally, the vibration spectrum (10-2000Hz) and temperature gradient (-40°C~85°C) analyzed in the MEMS sensor data can be used as environmental load data. The dynamic response relationship refers to the analysis of the structural characteristics of the package under the action of environmental loads, and the change of its physical, mechanical, electrical and other properties over time. For example, when the temperature changes, the internal structure of the package due to the difference in the thermal expansion coefficient of the material The above-mentioned failure modes refer to the identification of possible failures or performance degradation forms of the package under different working conditions based on the dynamic response relationship, including solder joint cracking, chip and substrate delamination, package shell rupture, electrical short circuit or open circuit, signal transmission distortion, etc. The above-mentioned damage evolution network refers to the model of the gradual development and evolution of package damage over time, environment, working conditions and other factors established for different failure modes. The above-mentioned parameterized geometric architecture refers to the package geometric model expressed in parameter form, which is achieved by defining a series of adjustable parameters (such as length, width, thickness, etc.). The shape and size of the package and its internal components are accurately described by using multi-source heterogeneous detection data (such as curvature, radius of curvature, angle, etc.). The physical characteristics refer to the deterministic mapping characteristics between the measurement data and the physical parameters of the entity object in the multi-source heterogeneous detection data, such as acoustic detection data (time-amplitude curve) → carries material elastic modulus information, and infrared thermal imaging data (temperature distribution) → reflects the thermal conductivity distribution. The multi-physics field attributes refer to the comprehensive attribute set obtained by assigning attribute parameters related to multiple physical fields such as mechanics, thermal, electrical, and acoustics to the parameterized geometric architecture. For example, the elastic modulus and Poisson's ratio of the material are assigned in the mechanical field; the thermal conductivity and thermal boundary conditions are set in the thermal field; and the conductivity and voltage boundary are defined in the electrical field.

[0030] Optionally, according to the dynamic response relationship, the failure mode of the package under different working conditions can be identified by a support vector machine model, and based on the failure mode, the damage evolution network of the package under the different working conditions can be set up using a finite element model, and the multi-physics field property assignment processing of the parameterized geometric architecture can be implemented using a multi-physics field coupling simulation platform.

[0031] Furthermore, the embodiment of the present invention inputs the packaging process parameters and material properties of the industrial solid-state hard drive into the virtual mirror body to generate a defect sensitivity distribution map of the package body, which can intuitively discover high-incidence areas of defects such as cold solder joints and chip delamination in the package body, quickly determine the detection focus, reduce unnecessary detection steps, and improve the defect detection efficiency of the industrial solid-state hard drive package body. The packaging process parameters refer to various key parameters used to control and describe the packaging manufacturing process, including thermal process parameters (such as reflow peak temperature, curing holding time), mechanical process parameters (such as patch pressure, molding holding pressure) and geometric process parameters (such as solder paste printing thickness, chip stacking offset). The material properties refer to the inherent physical, chemical, and mechanical property parameter sets exhibited by the component materials constituting the package body under specific environmental conditions, wherein the physical Properties include the material's density, thermal expansion coefficient, thermal conductivity, electrical conductivity, etc. For example, if the thermal expansion coefficient of the packaging substrate material does not match that of the chip material, thermal stress is likely to occur when the temperature changes, leading to delamination or cracks between the chip and the substrate. Chemical properties include the material's oxidation resistance and corrosion resistance. For example, some materials are prone to oxidation reactions in humid environments, which in turn affects the electrical properties of the package. Mechanical properties include the material's elastic modulus, yield strength, hardness, etc. For example, if the hardness and strength of the packaging shell material are insufficient, it may not be able to effectively protect the internal chip and circuit. The defect sensitivity distribution map refers to a visual chart generated based on a virtual mirror image of the industrial solid-state drive package, which is used to quantitatively characterize the tendency of the package to have specific types of defects at different locations. It is usually expressed as a dimensionless index between 0 and 1 or an actual physical quantity threshold.

[0032] As an embodiment of the present invention, inputting the packaging process parameters and material properties of the industrial solid-state drive into the virtual mirror body to generate a defect sensitivity distribution map of the package body includes: After inputting the packaging process parameters and material properties of the industrial solid-state drive into the virtual mirror body, identifying the energy distribution of the virtual mirror body; Calculating defect sensitivity indexes of different regions of the virtual mirror body based on the energy distribution; According to the defect sensitivity index, setting the defect sensitivity level of the different areas; Based on the defect sensitivity level, defining visualization color mapping rules for the different areas; Visualization processing of the different regions is performed according to the visualization color mapping rule to obtain a defect sensitivity distribution map of the package.

[0033] Among them, the energy distribution refers to the energy density stored or dissipated at different positions inside each region of the package when subjected to stress, including elastic strain energy, dissipated energy, surface energy, etc., which can be identified by the phase field method. For example, by inputting the energy distribution into the phase field simulation model, the crack propagation path and speed of the package can be predicted. The different regions refer to the solder ball array area, substrate wiring area, and chip mounting area divided according to the package structure. The defect sensitivity index refers to a dimensionless indicator that quantifies the probability of defects in each region of the package, which can be calculated based on stress-strain data and material failure threshold. The defect sensitivity level refers to the classification of different regions of the package according to the defect sensitivity index of each region. The defect risk of a domain is divided into several discrete risk levels, each level corresponding to a clear index range. For example, when the index is ≥0.7, it is set to a high risk level; when the index is 0.3≤<0.7, it is set to a medium risk level; when the index is <0.3, it is set to a low risk level. The visual color mapping rule refers to the correspondence between the defect sensitivity level and the color. Through this rule, the risk levels of different areas (such as high / medium / low) are mapped to specific colors in the visualization interface. The visualization processing refers to converting the defect sensitivity index of the virtual mirror body into an intuitive image or three-dimensional model through color filling, contour drawing, 3D cloud map, etc.

[0034] Optionally, the stress-strain distribution of the virtual mirror body can be discretized into a finite number of units (such as tetrahedrons and hexahedrons) and identified by solving the mechanical equations of each unit. The visualization color mapping rules of the different areas based on the defect sensitivity level can be defined using ANSYS simulation software.

[0035] S3. Collect multi-parameter time series data of the package in real time, identify the defect extension path of the package based on the multi-parameter time series data and the virtual mirror, and determine the defect type of the package according to the multi-dimensional defect characteristics.

[0036] The embodiment of the present invention can dynamically track the physical field changes of the entire solid-state industrial hard disk packaging process by real-time acquisition of multi-parameter time series data of the package body, accurately supporting multi-dimensional detection of packaging defects and failure tracing. The multi-parameter time series data refers to the sequence data of parameters such as vibration (mechanical stress), temperature (thermal stress), and humidity (environmental stress) that are acquired in real time by sensors and change over time, and is used to accurately restore multi-physical field coupling failures. For example, under high temperature (>85°C) + high humidity (>60%RH) + high-frequency vibration (>100Hz), the substrate resin absorbs moisture and expands + thermal stress + mechanical impact, which can easily cause substrate delamination defects.

[0037] Optionally, the multi-parameter time series data of the package body can be collected by multi-modal sensors, such as a three-axis acceleration sensor that can collect X / Y / Z axis vibration data; a thermocouple / infrared temperature sensor that can monitor heat distribution; and a temperature and humidity sensor that can record ambient and internal humidity.

[0038] Furthermore, the embodiments of the present invention can dynamically associate the causal relationship between environmental loads and defect evolution, thereby optimizing the design and material selection of industrial solid-state drive packages by identifying the defect expansion path of the package based on the multi-parameter time series data and the virtual mirror. The defect expansion path refers to the spatiotemporal evolution trajectory of internal defects in the package (such as solder ball microcracks and substrate delamination) from initial initiation to functional failure under the action of external stress (heat, mechanical, humidity, etc.).

[0039] As an embodiment of the present invention, identifying the defect extension path of the package based on the multi-parameter timing data and the virtual mirror includes: Loading the multi-parameter time series data through the virtual mirror body, performing multi-physics field coupling simulation processing on the package body, and obtaining a stress distribution simulation result; outputting key defect driving parameters of the package based on the stress distribution simulation results; Analyzing the defect expansion trend of the package body driven by the key defect driving parameters; defining an adaptive defect expansion threshold for the package; marking a critical defect expansion area of ​​the package according to the adaptive defect expansion threshold and the defect expansion trend; Extracting multi-parameter time series data features of the key defect area to construct a defect extension feature vector of the key defect area; Obtaining a defect extension path segment of the package based on the defect extension feature vector; The defect extension path segment is fitted by a path fitting algorithm to obtain a defect extension path.

[0040] Among them, the multi-physics field coupling simulation processing refers to the simulation processing that simultaneously considers the interaction and influence of multiple physical fields (such as thermal field, mechanical field, electric field, etc.) during the simulation process, including thermal-mechanical coupling and electro-thermal coupling, wherein the thermal-mechanical coupling solves the bidirectional interaction between the temperature field and the displacement field through a direct coupling method; the electro-thermal coupling uses the Joule thermal module to calculate the energy conversion between the current density field and the temperature field, the key defect driving parameters refer to parameters that have a significant impact on the formation and development of defects, including thermal fatigue index, crack driving force and interface degradation rate, the defect expansion trend refers to the potential expansion direction and speed of the defect under the action of the key defect driving parameters, the adaptive defect expansion threshold refers to the critical condition for determining the expansion of the package defect, including the warning threshold and the critical threshold, wherein the warning threshold is dynamically set through the risk assessment model according to the package design standard, historical failure data and reliability requirements; the critical threshold is determined based on the structural bearing capacity of the package and the functional failure criterion through mechanical performance testing and finite element analysis, and the key defect expansion The defect expansion region refers to an area marked as likely to undergo defect expansion based on an adaptive defect expansion threshold and defect expansion trend. The multi-parameter time series data feature refers to a feature quantity extracted from the multi-parameter time series data that can characterize the essential characteristics of defect expansion, such as the slope of the temperature curve reflecting the rate of thermal change, the frequency and amplitude of stress fluctuations reflecting the stability of the mechanical environment, and abnormal peaks in the current density distribution revealing local heating or electrical performance anomalies. The defect expansion feature vector refers to an ordered numerical vector integrated after normalization, dimensionality reduction, and encoding of the extracted multi-parameter time series data features. The defect expansion path segment refers to a basic unit that records the local process of defect expansion in the three-dimensional space of the package, where each segment must include the start / end point coordinates (x, y, z), an expansion velocity vector, and a material traversal sequence. The path fitting algorithm refers to an algorithm based on numerical analysis and curve fitting theory, including polynomial fitting, spline curve fitting, or least squares method. The fitting process refers to the process of optimizing and calculating discrete defect expansion path segments based on the selected path fitting algorithm.

[0041] Optionally, the multi-physics field coupling simulation processing of the package can be performed using a hybrid solver, the defect expansion trend of the package driven by the key defect driving parameters can be simulated by a phase field method, and the defect expansion path segments of the package based on the defect expansion eigenvector can be obtained using a Delaunay triangulation algorithm.

[0042] By determining the defect type of the package based on the multi-dimensional defect characteristics, the embodiments of the present invention can transform the defects of the industrial solid-state drive package from fuzzy judgment to precise classification, thereby improving the targeted detection of industrial solid-state drive package defects. The defect type refers to the abnormal state of the industrial solid-state drive package caused by material, process or structural problems, and is standardized according to its manifestation, cause, impact and other characteristics, such as solder joint cracking, chip displacement, package delamination, and excessive voids.

[0043] As an embodiment of the present invention, determining the defect type of the package according to the multi-dimensional defect characteristics includes: Identifying macroscopic structural features, mesoscopic microstructural unit features, and microstructural features of the package from the multidimensional defect features; performing cross-scale correlation processing of the macroscopic structural features, the mesoscopic microstructure unit features, and the microstructure features to obtain a fused defect feature list; Calculating the feature importance coefficient and cross-scale defect correlation of each feature in the fused defect feature list; Extracting core defect features of the package based on the feature importance coefficient and the cross-scale defect correlation; The defect type of the package is determined based on the core defect characteristics.

[0044] Among them, the macroscopic structural features refer to the geometric morphology and physical properties of the package at the millimeter level (10⁻³m) or above, such as solder ball height deviation, package warpage, pin coplanarity, crack length, etc., which can be obtained by laser three-dimensional profilometer. The mesoscopic microstructure unit features refer to the micron level (10⁻³m) inside the package. 6 m) Structural characteristics of functional units, such as the thickness of IMC (intermetallic compound) inside the solder joint, the curvature of the bonding wire, the void ratio of the chip-substrate interface, etc., can be extracted using SEM microscopic imaging. The microstructural features refer to the nanoscale (10⁻ 9m) Material crystal structure and surface properties, such as grain orientation, dislocation density, interface atomic diffusivity, and surface energy distribution, can be obtained using an atomic force microscope (AFM). Cross-scale correlation processing refers to the process of establishing a quantitative relationship between the package's macrostructure (such as overall deformation), mesoscopic microstructure (such as solder joint grains), and microscopic features (such as dislocations and interfaces) through an algorithm. The feature importance coefficient is an indicator that measures the contribution of a single feature to the determination of the defect type and is typically normalized to the interval [0,1]. The cross-scale defect correlation refers to the nonlinear coupling effect between features of different scales. The core defect feature refers to the most discriminative feature subset selected from the fused features. The core defect feature must meet the following requirements: a SHAP value ≥ 0.1 (importance threshold), an absolute value of the cross-scale defect correlation > 0.5, and be located within a 50μm buffer zone at the material interface.

[0045] Optionally, the cross-scale association processing of the macroscopic structural features, the mesoscopic microstructure unit features and the microstructure features can be implemented using a graph attention fusion network algorithm, and the feature importance coefficients of various features in the fused defect feature list can be calculated by a SHAP algorithm. Based on the core defect features, the defect types of the package can be matched by establishing a multimodal knowledge graph. For example, a knowledge graph can be constructed with defect types (voids / cracks / delamination, etc.) as nodes and association rules (such as importance threshold ≥ 0.1) as edges, and a graph neural network (GNN) algorithm can be used to calculate the similarity between the core defect features and the defect types.

[0046] In an optional embodiment of the present invention, the cross-scale defect correlation of each type of feature in the fused defect feature list is calculated using the following formula: ; in, Indicates the cross-scale defect correlation of various features in the fusion defect feature list, Represents the scale index in the fusion defect feature list, 3 represents the macroscopic, mesoscopic and microscopic scales in the fusion defect feature list, represents the scale weight, represents the macroscopic strain tensor in the fused defect feature list, represents the mesoscopic grain orientation distribution matrix in the fused defect feature list, represents the microscopic dislocation density tensor in the fused defect feature list, represents the trace of the matrix, T represents the transpose of the matrix, Represents the determinant of a matrix.

[0047] It should be noted that in this application, the macro-meso-scopic three-scale physical quantities are converted into Direct correlation is used for cross-scale coupling of multimodal defect features. In particular, it should be noted that by introducing The exponential structure can quantify the nonlinear synergistic effects between multimodal defect features. For example, when m = 1, the macro-dominant effect is strengthened, and when m = 3, the micro-nonlinear mutation is captured.

[0048] S4. Locate process weaknesses of the package according to the defect type and the defect sensitivity distribution map, and generate a graded defect detection instruction for the package based on the process weaknesses, the defect extension path, and the defect type.

[0049] The embodiment of the present invention locates the process weak links of the package body according to the defect type and the defect sensitivity distribution map, and can quickly focus on the key risk points of the industrial solid-state hard drive packaging process to optimize the corresponding process parameter settings, thereby effectively reducing the defect rate in the industrial solid-state hard drive packaging process and improving the overall product quality and production efficiency. The process weak links refer to the key processes or parameter combinations in the packaging production process where defects are more likely to occur and have a significant impact on product quality due to factors such as process design, parameter settings, equipment accuracy or material compatibility. For example, if the chip offset defect accounts for more than 30% in a certain packaging process, and a slight fluctuation in its corresponding process parameters (such as patch pressure) will significantly increase the defect incidence rate, then this process is a process weak link.

[0050] As an embodiment of the present invention, locating a process weak link of the package body according to the defect type and the defect sensitivity distribution map includes: Calculate the occurrence frequency of the defect type, and filter out high-frequency defect categories based on the occurrence frequency; Identifying a defect location corresponding to the high-frequency defect category and obtaining point cloud data of the defect location; Performing spatial registration processing on the point cloud data and the defect sensitivity distribution map to obtain a spatial registration result; Based on the spatial registration result, identifying an overlapping area between the point cloud data and the defect sensitivity distribution map; Retrieving historical process parameter data corresponding to the package according to the overlapping area; Analyzing the correlation between the process parameter historical data and the defect type; Based on the association relationship, identifying key process steps and operating parameters corresponding to the defect type; The process weaknesses of the package are located through the key process steps and operating parameters.

[0051] Among them, the occurrence frequency refers to the proportion of the number of times a certain type of defect occurs in a certain production cycle or total sample volume to the total number of defects. For example, if 1,000 packages are counted and the solder joint cracking defect occurs 50 times, then its occurrence frequency is 5%. The high-frequency defect category refers to a set of defect types whose occurrence frequency is higher than a set threshold (such as 3%). If the critical threshold of the occurrence frequency is set to 2%, then solder joint cracking (5%) and package bubbles (4%) can be classified as high-frequency defects. The defect location refers to the physical coordinates or area where the defect actually occurs on the package, and the precise location information can be obtained through X-ray, CT scanning and other equipment. The point cloud data refers to a spatial coordinate set composed of a large number of discrete points collected by three-dimensional scanning equipment (such as lidar, industrial CT), which is used to characterize the geometric characteristics of the defect location. The spatial registration processing refers to the process of aligning the coordinates and unifying the scale of the point cloud data with the defect sensitivity distribution map. The process can be performed by mapping the three-dimensional coordinates of the point cloud to the two-dimensional plane or three-dimensional space of the sensitivity distribution map through feature matching algorithms (such as the iterative closest point algorithm ICP), affine transformation, etc. The overlapping area refers to the part where the point cloud data (actual position of the defect) overlaps with the high-sensitivity area in the defect sensitivity distribution map after spatial alignment. The process parameter historical data refers to the historical records of parameter settings and actual operation data of each process step in the packaging production process, such as temperature, pressure, time, speed, etc. The correlation relationship refers to the causal or probabilistic relationship between the process parameters and the defect type. The key process step refers to the production process that has a decisive influence on the occurrence of a specific defect type. For example, if the analysis shows that solder joint cracking is strongly correlated with the reflow soldering temperature curve, the reflow soldering process is a key process step. The operating condition parameters refer to specific parameters in the key process steps that have a significant impact on defect formation, such as temperature, pressure, speed, etc.

[0052] Optionally, based on the overlapping area, the historical data of process parameters corresponding to the package body can be matched through a spatiotemporal indexing method, and the correlation between the historical data of process parameters and the defect type can be analyzed using a causal reasoning model, such as a Bayesian network model. Based on the correlation, the key process steps and operating parameters corresponding to the defect type can be identified through an LSTM-AE model.

[0053] Furthermore, the embodiment of the present invention generates graded defect detection instructions for the package body by combining the process weaknesses, the defect extension paths and the defect types, thereby effectively integrating detection resources and avoiding efficiency loss caused by blind detection. The graded defect detection instructions refer to targeted detection plans generated after risk level classification of the package body detection tasks based on process weaknesses and defect extension paths.

[0054] As an embodiment of the present invention, the step of generating a hierarchical defect detection instruction for the package body in combination with the process weak link, the defect extension path, and the defect type includes: Identify the abnormal frequency of process parameters corresponding to the weak links in the process; Calculating the probability of defect occurrence in each area of ​​the package body according to the abnormal frequency of the process parameters; identifying a defect diffusion rate of the package body in the defect propagation path; Determining the degree of damage caused by the defect of the package according to the defect type; Establishing a three-dimensional risk matrix for the package body according to the defect occurrence probability, the defect diffusion speed, and the defect hazard degree; determining a defect risk level of the package based on the three-dimensional risk matrix; Develop an executable detection protocol corresponding to the defect risk level; A graded defect inspection instruction for the package is generated according to the defect risk level and the executable inspection protocol.

[0055] Among them, the frequency of process parameter abnormality refers to the number of times the process parameters in the packaging production process deviate from the preset standard range within the statistical period, for example, the number of times the reflow soldering temperature exceeds the [230℃-240℃] range, or the number of times the patch pressure is lower than the lower limit. The probability of defect occurrence refers to the possibility of specific defects occurring in various areas of the package body, which can be obtained through Monte Carlo simulation. The defect diffusion rate refers to the rate at which the defect expands from the initial position to the surrounding area, and the unit is usually μm / s (micro defects) or mm / h (macro defects), which can be determined through accelerated aging tests. The degree of defect hazard refers to a quantitative description of the serious consequences of the defect based on factors such as the size, location, frequency of occurrence and potential destructive ability to product performance corresponding to the defect type. For example, a cold solder joint on the main control chip will directly lead to data transmission interruption, and the hard disk will not be able to operate normally. work, such defects have a high degree of hazard; while slight scratches on the surface of the package shell only affect the appearance, do not affect the performance, and have a low degree of hazard. The three-dimensional risk matrix refers to a three-dimensional assessment model constructed with defect occurrence probability, defect diffusion speed and defect hazard degree as three dimensions, which is used to comprehensively quantify the risk level of each area of ​​the package. The defect risk level refers to the risk level of each area of ​​the package divided according to the calculation results of the three-dimensional risk matrix, which is usually divided into high risk (immediate detection and treatment is required), medium risk (routine sampling), and low risk (regular monitoring). The executable detection protocol refers to a standardized detection plan formulated for different defect risk levels, including specific operating details such as detection equipment selection (such as CT, SEM), detection frequency (such as once an hour in high-risk areas), detection accuracy (such as nanometer-level resolution), and detection process (such as macro first and then micro).

[0056] S5. Query the hardware configuration of the industrial solid-state drive, extract the number of NAND stacking layers and the solder ball pitch in the hardware configuration, and set a personalized defect detection method for the package body based on the number of NAND stacking layers, the solder ball pitch, and the defect type.

[0057] The embodiment of the present invention queries the hardware configuration of the industrial solid-state drive and extracts the number of NAND stacking layers and solder ball pitch in the hardware configuration. It can adjust the defect detection strategy in a targeted manner based on the characteristics of different numbers of stacking layers and solder ball pitches, thereby significantly improving the detection efficiency and the accuracy of defect identification. The hardware configuration refers to the collection of all physical components and their parameter settings inside the industrial solid-state drive, including the NAND flash memory chip type, main control chip model, cache capacity, circuit board layout, packaging form, etc. The number of NAND stacking layers refers to the number of layers of NAND flash memory units stacked in the vertical direction in the NAND flash memory manufacturing process. The solder ball pitch refers to the center distance between the solder balls used to connect the chip to the circuit board in flip-chip packaging technology. The unit is usually millimeter (mm) or micron (μm). It is one of the key parameters for measuring packaging accuracy and reliability. The smaller the pitch, the more solder balls can be arranged per unit area.

[0058] Furthermore, the embodiment of the present invention sets a personalized defect detection method for the package body by combining the number of NAND stacking layers, the solder ball pitch and the defect type, thereby achieving deep adaptation of the detection strategy with hardware characteristics and defect characteristics, accurately locating process risk points, and significantly improving the efficiency of package defect detection. The personalized defect detection method refers to a customized detection scheme dynamically generated based on the specific structural parameters of the package body (number of NAND stacking layers, solder ball pitch) and defect characteristics, such as generating a detection instruction set including a detection path, parameter settings and execution timing, and transmitting it to the detection equipment.

[0059] As an embodiment of the present invention, the personalized defect detection method for the package is set in combination with the number of NAND stacking layers, the solder ball pitch, and the defect type, including: Extracting the interlayer interconnection density and thermal stress distribution characteristics of the package according to the number of NAND stack layers; Calculating the micro solder joint detection accuracy of the package body according to the solder ball spacing; Identifying a key inspection area of ​​the package body based on the interlayer interconnection density, the thermal stress distribution characteristics, and the micro-solder point inspection accuracy; Based on the defect type, analyzing the defect formation mechanism and detection sensitivity requirements of the package; Establishing a defect feature database of the package using the defect formation mechanism and the detection sensitivity requirement; Creating a multimodal collaborative inspection scheme for the package according to the key inspection area and the defect feature database; Identify the detection modalities corresponding to the multimodal collaborative detection solution and define the timing coordination rules between the detection modalities; According to the multimodal collaborative detection scheme and the timing coordination rule, a personalized defect detection method for the package body is set.

[0060] To better understand the extraction process of interlayer interconnection density and thermal distribution characteristics, refer to Figure 2 The figure shows a schematic diagram of the NAND stacking structure, which presents the layered physical architecture of the molding compound, upper chip, lower chip, chip mounting film and substrate. Through this structure, the vertical / horizontal layout of the inter-layer interconnection under different stacking layers can be clearly determined (for example, the interconnection line needs to pass through the chip mounting film to connect the upper and lower chips), thereby assisting in calculating the inter-layer interconnection density; at the same time, based on the different material layers in the figure (such as the difference in thermal expansion coefficient between the molding compound and the chip), the transmission path of thermal stress at the chip-film-substrate interface can be analyzed to assist in the extraction of thermal stress distribution characteristics.

[0061] Among them, the inter-layer interconnection density refers to the number of interconnection structures such as connecting lines and vias between different storage layers per unit area in the NAND stacking structure, which can be obtained through finite element analysis. For example, for a 128-layer stacked NAND, the density in the central area is measured to be ≥5000 / mm², and the density in the edge area is ≤3000 / mm². The thermal stress distribution characteristics refer to the size, direction and distribution pattern of thermal stress generated in different areas of the NAND stacking package due to factors such as differences in material thermal expansion coefficients and working heat, which can be obtained through thermal imaging detection. The micro-solder point detection accuracy refers to the ability of the detection equipment to distinguish and measure the minimum physical size or defect feature when detecting micro-solder points of the package, usually expressed in length units (such as microns μm). For example, if the detection equipment can identify solder joint voids with a diameter of 0.1μm, its detection accuracy is 0.1μm. The key inspection area refers to the area in the package that is most prone to defects and has the greatest impact on product performance, determined through comprehensive analysis based on parameters such as interlayer interconnection density, thermal stress distribution characteristics, and micro-solder joint detection accuracy. The key inspection area includes the stacking layer mutation area (such as the transition area from 64 layers to 128 layers), the thermal stress concentration area (the interface with CTE difference > 2ppm / °C), and the solder ball array density mutation area (the transition zone with spacing change > 15%). The defect formation mechanism refers to the cause, physical and chemical process, and influencing factors of a specific type of defect during the packaging process. For example, solder joint cold soldering is caused by factors such as insufficient soldering temperature, solder ball oxidation, and pad contamination. The detection sensitivity requirement refers to the performance requirements for the detection equipment and method in terms of minimum detectable defect size, signal recognition capability, etc., based on the defect type and formation mechanism, to ensure the effective detection of the target defect. For example, delamination defects require ultrasonic detection SNR ≥ 15dB; micro voids require X-ray CT contrast ≥ 8%. The defect feature database refers to the defects generated by the ultrasonic detection SNR ≥ 15dB; micro voids require X-ray CT contrast ≥ 8%. By collecting, organizing, and analyzing a large amount of historical inspection data, a database is established that contains information such as morphological characteristics (such as void diameter and crack orientation angle) of different defect types, signal characteristics (such as ultrasonic echo spectrum peak), and environmental parameters (such as vacuum fluctuation during bonding). The multimodal collaborative inspection solution refers to integrating the advantages of multiple inspection technologies (such as X-ray inspection, electron microscopy inspection, ultrasonic inspection, thermal imaging inspection, etc.) to develop a collaborative working inspection strategy for key inspection areas and different defect types of the package. The inspection modality refers to a specific inspection technology or method, and each inspection modality has a unique inspection principle and scope of application. For example, the X-ray inspection modality uses the penetration of X-rays to detect internal structural defects, and the electron microscopy inspection modality is used to observe microscopic surface morphology. The timing coordination rules refer to the rules that specify the execution order, time interval, and data interaction logic of each inspection modality in the multimodal inspection solution. For example, the minimum time interval between X-ray exposure and optical scanning is ≥50ms; the phase difference between ultrasonic pulse emission and electron beam scanning is controlled at ±5μs.

[0062] Optionally, the micro solder joint detection accuracy of the package body according to the solder ball pitch can be calculated based on the ratio of the solder ball pitch to the minimum detectable defect size, the defect formation mechanism of the package body based on the defect type can be determined by failure mode analysis (FMEA), and the detection sensitivity requirement can be determined using a defect size distribution curve.

[0063] S6. Combining the hierarchical defect detection instruction and the personalized defect detection method, outputting the defect detection result of the package through the virtual mirror body.

[0064] The embodiment of the present invention combines the hierarchical defect detection instructions and the personalized defect detection method, and outputs the defect detection results of the package through the virtual mirror body. This not only enables the precise allocation of detection resources according to the defect risk level and package characteristics, significantly improving the defect detection rate in high-risk areas, but also enables the use of the virtual mirror body to simulate and verify the detection process and results in real time, predict detection blind spots and potential errors in advance, achieve dynamic optimization of the detection strategy, and effectively improve the accuracy and reliability of the detection results. The defect detection result refers to the comprehensive information about the defect status obtained after the package is inspected by executing the personalized defect detection method, including defect existence determination, defect type and characteristics, defect severity assessment, impact analysis and risk prediction results. For example, the output result may be that a delamination defect exists between the NAND stacking layers of the industrial solid-state drive package and is located in the 15th-20th layer area.

[0065] Exemplarily, in combination with the hierarchical defect detection instruction and the personalized defect detection method, the specific steps for outputting the defect detection results of the package through the virtual mirror body are: parsing the risk level division, detection priority and standards in the hierarchical defect detection instruction, and disassembling the multimodal scheme, parameters and scheduling rules of the personalized defect detection method; then, inputting the package design parameters, process data and historical defect data based on the detection instruction and the detection method, and configuring the simulation accuracy and data parsing rules of the virtual mirror body; then, performing multimodal detection such as 3D X-ray and electron microscope according to the timing rules of personalized detection, and synchronizing the real-time data to the virtual mirror body, and then using the virtual mirror body to simulate the detection process, compare the actual and simulated results, verify the detection parameters and predict the defect evolution trend; then, fusing the multimodal detection raw data and the virtual mirror body analysis results to identify defect characteristics; finally, generating a three-dimensional defect map in the virtual mirror body, outputting a report containing detection and analysis data and defect prediction, and feeding it back to the production system, thereby realizing the output of package defect detection results based on the virtual mirror body.

[0066] Compared with the problems described in the background technology, the embodiment of the present invention collects multi-source heterogeneous detection data of the package body and extracts multi-dimensional defect characteristics of the package body based on the multi-source heterogeneous detection data, which can provide data support for the optimization of the industrial solid-state hard disk packaging process; further, the embodiment of the present invention constructs a virtual mirror body of the package body by utilizing the multi-source heterogeneous detection data, which can accurately locate the weak links in the process and formulate personalized detection strategies, thereby effectively improving the pertinence and accuracy of the defect detection of the industrial solid-state hard disk package body. The embodiment of the present invention inputs the packaging process parameters and material properties of the industrial solid-state hard disk into the virtual mirror body The invention can generate a defect sensitivity distribution map of the package body, and can intuitively find high-incidence areas of defects such as cold solder joints and chip delamination in the package body, quickly determine the detection focus, reduce unnecessary detection steps, and improve the defect detection efficiency of the industrial solid-state hard disk package body; the embodiment of the present invention can identify the defect expansion path of the package body based on the multi-parameter time series data and the virtual mirror body, and can dynamically associate the causal relationship between environmental load and defect evolution, optimize the design and material selection of the industrial solid-state hard disk package body, and further, the embodiment of the present invention can determine the defect type of the package body according to the multi-dimensional defect characteristics, and can identify the defect type of the industrial solid-state hard disk package body. The defects are converted from fuzzy judgment to precise classification, thereby improving the pertinence of defect detection of industrial solid-state hard disk packages; the embodiment of the present invention locates the process weaknesses of the package according to the defect type and the defect sensitivity distribution map, and can quickly focus on the key risk points of the industrial solid-state hard disk packaging process to optimize the corresponding process parameter settings, thereby effectively reducing the defect rate in the industrial solid-state hard disk packaging process and improving the overall product quality and production efficiency. Furthermore, the embodiment of the present invention generates a hierarchical defect detection instruction for the package by combining the process weaknesses, the defect extension path and the defect type, which can effectively integrate detection resources. Avoid efficiency loss caused by blind detection; the embodiment of the present invention queries the hardware configuration of the industrial solid-state drive and extracts the number of NAND stacking layers and solder ball pitch in the hardware configuration. It can adjust the defect detection strategy in a targeted manner according to the characteristics of different numbers of stacking layers and solder ball pitches, significantly improving the detection efficiency and the accuracy of defect identification. Furthermore, the embodiment of the present invention sets a personalized defect detection method for the package body by combining the number of NAND stacking layers, the solder ball pitch and the defect type. It can achieve deep adaptation of the detection strategy with hardware characteristics and defect characteristics, accurately locate process risk points, and significantly improve the efficiency of package defect detection.Finally, the embodiment of the present invention combines the hierarchical defect detection instructions with the personalized defect detection method, and outputs the defect detection results of the package through the virtual mirror. This not only enables the precise allocation of detection resources according to the defect risk level and package characteristics, significantly improving the defect detection rate in high-risk areas, but also enables the use of the virtual mirror to simulate and verify the detection process and results in real time, predict detection blind spots and potential errors in advance, achieve dynamic optimization of detection strategies, and effectively improve the accuracy and reliability of detection results. Therefore, the embodiment of the present invention provides a method and system for industrial solid-state drive package defect detection, which can effectively improve the accuracy and reliability of industrial solid-state drive package defect detection results.

[0067] Example 2: like Figure 3 FIG. 1 is a functional module diagram of a packaging defect detection system for an industrial solid-state hard drive according to the present invention.

[0068] The industrial solid-state drive packaging defect detection system 200 described in the present invention can be installed in an electronic device. Depending on the functionality implemented, the industrial solid-state drive packaging defect detection system may include a data acquisition module 201, a virtual image construction module 202, a defect analysis module 203, a detection instruction generation module 204, an execution mode determination module 205, and a result output module 206. A module, also referred to as a unit, is a series of computer program segments that can be executed by an electronic device processor and perform a fixed function, and is stored in the electronic device's memory.

[0069] In the embodiment of the present invention, the functions of each module / unit are as follows: The data acquisition module 201 is configured to acquire a package of an industrial solid-state drive, collect multi-source heterogeneous detection data of the package, and extract multi-dimensional defect features of the package based on the multi-source heterogeneous detection data; The virtual mirror body construction module 202 is configured to construct a virtual mirror body of the package body using the multi-source heterogeneous detection data, and input the packaging process parameters and material properties of the industrial solid-state drive into the virtual mirror body to generate a defect sensitivity distribution map of the package body; The defect analysis module 203 is configured to collect multi-parameter time series data of the package in real time, identify the defect extension path of the package based on the multi-parameter time series data and the virtual mirror, and determine the defect type of the package according to the multi-dimensional defect characteristics; The detection instruction generation module 204 is configured to locate the process weakness of the package according to the defect type and the defect sensitivity distribution map, and generate a graded defect detection instruction for the package based on the process weakness, the defect extension path, and the defect type; The execution mode determination module 205 is used to query the hardware configuration of the industrial solid-state drive, extract the number of NAND stacking layers and the solder ball pitch in the hardware configuration, and set a personalized defect detection mode for the package body based on the number of NAND stacking layers, the solder ball pitch, and the defect type; The result output module 206 is configured to combine the graded defect detection instruction and the personalized defect detection method to output the defect detection result of the package through the virtual mirror.

[0070] In detail, the modules in the industrial solid state drive packaging defect detection system 200 described in the embodiment of the present invention are used in the same manner as described above. Figure 1 The same technical means are used as the method for detecting packaging defects of industrial solid-state drives described in , and can produce the same technical effects, so they will not be repeated here.

[0071] 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.

[0072] 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 detecting packaging defects of industrial solid-state hard drives, characterized in that: The method comprises: Obtaining a package of an industrial solid-state drive, collecting multi-source heterogeneous detection data of the package, and extracting multi-dimensional defect features of the package based on the multi-source heterogeneous detection data; Constructing a virtual mirror image of the package using the multi-source heterogeneous detection data, and inputting the packaging process parameters and material properties of the industrial solid-state drive into the virtual mirror image to generate a defect sensitivity distribution map of the package; collecting multi-parameter time series data of the package in real time, identifying a defect extension path of the package based on the multi-parameter time series data and the virtual mirror, and determining a defect type of the package according to the multi-dimensional defect characteristics; Locating process weaknesses of the package according to the defect type and the defect sensitivity distribution map, and generating a graded defect detection instruction for the package based on the process weaknesses, the defect extension path, and the defect type; querying the hardware configuration of the industrial solid-state drive, extracting the number of NAND stacking layers and the solder ball pitch in the hardware configuration, and setting a personalized defect detection method for the package body based on the number of NAND stacking layers, the solder ball pitch, and the defect type; In combination with the hierarchical defect detection instruction and the personalized defect detection method, the defect detection result of the package body is output through the virtual mirror body.

2. The method for detecting packaging defects of an industrial solid-state drive according to claim 1, wherein: The extracting multi-dimensional defect features of the package based on the multi-source heterogeneous detection data includes: Performing spatiotemporal registration processing on the multi-source heterogeneous detection data to obtain target fusion data; extracting image data, electrical parameter data, acoustic image data, and environmental response data from the target fusion data; performing image enhancement processing on the image data to obtain enhanced image data; Segmenting a defect area from the enhanced image data and extracting defect geometric features of the defect area; Performing outlier elimination and smoothing processing on the electrical parameter data to obtain a target electrical parameter sequence; Analyzing parameter change trends in the target electrical parameter sequence to extract abnormal electrical fluctuation characteristics in the target electrical parameter sequence; Performing variational modal decomposition processing on the ultrasonic signal in the acoustic image data to obtain a characteristic frequency band, and calculating a frequency band energy ratio of the characteristic frequency band; extracting environmental sensitivity characteristics of the package based on the environmental response data; The multi-dimensional defect characteristics of the package are determined based on the defect geometric characteristics, the abnormal electrical fluctuation characteristics, the frequency band energy proportion and the environmental sensitivity characteristics.

3. The method for detecting packaging defects of an industrial solid-state drive according to claim 1, wherein: The method of constructing a virtual mirror image of the package body by using the multi-source heterogeneous detection data includes: Determining structural features of the package and corresponding material states thereof through the multi-source heterogeneous detection data, and identifying environmental load data of the package; Analyzing the dynamic response relationship between the structural characteristics and the environmental load data; identifying failure modes of the package under different working conditions according to the dynamic response relationship; Based on the failure mode, setting a damage evolution network of the package under the different working conditions; constructing a parameterized geometric structure of the package according to the structural features and the material state; Performing multi-physics field attribute assignment processing on the parameterized geometric architecture based on the physical characteristics of the multi-source heterogeneous detection data to obtain multi-physics field attributes; Combining the damage evolution network, the parameterized geometric framework and the multi-physics field properties, a virtual mirror body of the package is constructed.

4. The method for detecting packaging defects of an industrial solid-state drive according to claim 1, wherein: Inputting the packaging process parameters and material properties of the industrial solid-state drive into the virtual mirror body to generate a defect sensitivity distribution map of the package body includes: After inputting the packaging process parameters and material properties of the industrial solid-state drive into the virtual mirror body, identifying the energy distribution of the virtual mirror body; Calculating defect sensitivity indexes of different regions of the virtual mirror body based on the energy distribution; According to the defect sensitivity index, setting the defect sensitivity level of the different areas; Based on the defect sensitivity level, defining visualization color mapping rules for the different areas; Visualization processing of the different regions is performed according to the visualization color mapping rule to obtain a defect sensitivity distribution map of the package.

5. The method for detecting packaging defects of an industrial solid-state drive according to claim 1, wherein: The identifying the defect extension path of the package based on the multi-parameter timing data and the virtual mirror body includes: Loading the multi-parameter time series data through the virtual mirror body, performing multi-physics field coupling simulation processing on the package body, and obtaining a stress distribution simulation result; outputting key defect driving parameters of the package based on the stress distribution simulation results; Analyzing the defect expansion trend of the package body driven by the key defect driving parameters; defining an adaptive defect expansion threshold for the package; marking a critical defect expansion area of ​​the package according to the adaptive defect expansion threshold and the defect expansion trend; Extracting multi-parameter time series data features of the key defect area to construct a defect extension feature vector of the key defect area; Obtaining a defect extension path segment of the package based on the defect extension feature vector; The defect extension path segment is fitted by a path fitting algorithm to obtain a defect extension path.

6. The method for detecting packaging defects of an industrial solid-state drive according to claim 1, wherein: The step of determining the defect type of the package according to the multi-dimensional defect characteristics includes: Identifying macroscopic structural features, mesoscopic microstructural unit features, and microstructural features of the package from the multidimensional defect features; performing cross-scale correlation processing of the macroscopic structural features, the mesoscopic microstructure unit features, and the microstructure features to obtain a fused defect feature list; Calculating the feature importance coefficient and cross-scale defect correlation of each feature in the fused defect feature list; Extracting core defect features of the package based on the feature importance coefficient and the cross-scale defect correlation; The defect type of the package is determined based on the core defect characteristics.

7. The method for detecting packaging defects of an industrial solid-state drive according to claim 1, wherein: The step of locating a process weak link of the package body according to the defect type and the defect sensitivity distribution map includes: Calculate the occurrence frequency of the defect type, and filter out high-frequency defect categories based on the occurrence frequency; Identifying a defect location corresponding to the high-frequency defect category and obtaining point cloud data of the defect location; Performing spatial registration processing on the point cloud data and the defect sensitivity distribution map to obtain a spatial registration result; Based on the spatial registration result, identifying an overlapping area between the point cloud data and the defect sensitivity distribution map; Retrieving historical process parameter data corresponding to the package according to the overlapping area; Analyzing the correlation between the process parameter historical data and the defect type; Based on the association relationship, identifying key process steps and operating parameters corresponding to the defect type; The process weaknesses of the package are located through the key process steps and operating parameters.

8. The method for detecting packaging defects of an industrial solid-state drive according to claim 1, wherein: The step of generating a graded defect detection instruction for the package body by combining the process weak link, the defect extension path, and the defect type includes: Identify the abnormal frequency of process parameters corresponding to the weak links in the process; Calculating the probability of defect occurrence in each area of ​​the package body according to the abnormal frequency of the process parameters; identifying a defect diffusion rate of the package body in the defect propagation path; Determining the degree of damage caused by the defect of the package according to the defect type; Establishing a three-dimensional risk matrix for the package body according to the defect occurrence probability, the defect diffusion speed, and the defect hazard degree; determining a defect risk level of the package based on the three-dimensional risk matrix; Develop an executable detection protocol corresponding to the defect risk level; A graded defect inspection instruction for the package is generated according to the defect risk level and the executable inspection protocol.

9. The method for detecting packaging defects of an industrial solid-state drive according to claim 1, wherein: The step of setting a personalized defect detection method for the package body in combination with the number of NAND stacking layers, the solder ball pitch, and the defect type includes: Extracting the interlayer interconnection density and thermal stress distribution characteristics of the package according to the number of NAND stack layers; Calculating the micro solder joint detection accuracy of the package body according to the solder ball spacing; Identifying a key inspection area of ​​the package body based on the interlayer interconnection density, the thermal stress distribution characteristics, and the micro-solder point inspection accuracy; Based on the defect type, analyzing the defect formation mechanism and detection sensitivity requirements of the package; Establishing a defect feature database of the package using the defect formation mechanism and the detection sensitivity requirement; Creating a multimodal collaborative inspection scheme for the package according to the key inspection area and the defect feature database; Identify the detection modalities corresponding to the multimodal collaborative detection solution and define the timing coordination rules between the detection modalities; According to the multimodal collaborative detection scheme and the timing coordination rule, a personalized defect detection method for the package body is set.

10. A packaging defect detection system for industrial solid-state hard drives, characterized in that: The system comprises: a data acquisition module, configured to acquire a package of an industrial solid-state drive, collect multi-source heterogeneous detection data of the package, and extract multi-dimensional defect features of the package based on the multi-source heterogeneous detection data; a virtual mirror body construction module, configured to construct a virtual mirror body of the package body using the multi-source heterogeneous detection data, and input the packaging process parameters and material properties of the industrial solid-state drive into the virtual mirror body to generate a defect sensitivity distribution map of the package body; a defect analysis module configured to collect multi-parameter time series data of the package in real time, identify a defect extension path of the package based on the multi-parameter time series data and the virtual mirror, and determine a defect type of the package based on the multi-dimensional defect characteristics; a detection instruction generation module, configured to locate a process weakness of the package according to the defect type and the defect sensitivity distribution map, and generate a graded defect detection instruction for the package based on the process weakness, the defect extension path, and the defect type; an execution mode determination module, configured to query the hardware configuration of the industrial solid-state drive, extract the number of NAND stacking layers and the solder ball pitch in the hardware configuration, and set a personalized defect detection mode for the package body based on the number of NAND stacking layers, the solder ball pitch, and the defect type; A result output module is configured to combine the graded defect detection instruction and the personalized defect detection method to output the defect detection result of the package through the virtual mirror body.

Images