Defect detection method and system for intelligent card chip based on deep learning

Abstract

The invention relates to the technical field of semiconductor chip manufacturing and detection, and discloses a method and system for realizing defect detection of an intelligent card chip based on deep learning, and the method comprises the steps: collecting a thermoluminescence signal of the surface of the intelligent card chip through a thermoluminescence detector, and generating a defect thermal topological graph of the surface of the intelligent card chip through the thermoluminescence signal; according to the defect thermal topological graph, carrying out phonon backstepping processing on the interior of the intelligent card chip to obtain phonon angular frequency of the interior of the intelligent card chip; based on the phonon angular frequency, constructing a phonon vibration frequency and virtual Joule heat distribution diagram of the smart card chip; according to the phonon vibration frequency and the virtual Joule heat distribution diagram, injecting modulation heat waves into the intelligent card chip so as to acquire in-phase heat wave data and anti-phase heat wave data of the intelligent card chip in response to the modulation heat waves; and identifying a defect area detection result of the intelligent card chip through the in-phase thermal wave data and the anti-phase thermal wave data. According to the invention, the defect detection accuracy of the existing chip can be improved.

Description

Technical Field

[0001] The present invention relates to a method and system for implementing defect detection of smart card chips based on deep learning, belonging to the technical field of semiconductor chip manufacturing and detection. Background Art

[0002] Nowadays, with the widespread application of smart cards in finance, transportation, identity authentication and other fields, chip quality is directly related to data security and system reliability. However, traditional chip defect detection technology still has significant limitations.

[0003] At present, the mainstream detection methods are still based on manual visual inspection or basic electrical signal testing. For example, in the chip mounting process, quality inspectors need to inspect multiple chips per minute, which is prone to visual fatigue under high-intensity work, resulting in a high rate of missed detection of micron-level defects such as lattice distortion and metal migration in manual inspection, and are unable to identify hidden defects such as finer tilt angles or bubbles in the glue layer. Second, existing automated inspection equipment mostly relies on a single signal parameter and does not integrate multi-dimensional features. For example, existing technologies only analyze electromagnetic radiation or power signals, and traditional machine vision algorithms only extract surface geometric features, ignoring the causal relationship between electrical signals and physical defects. This results in only reporting "abnormalities" but not being able to locate the defect position, and the fault location error is high.

[0004] Therefore, the accuracy of existing chip defect detection is insufficient. Summary of the Invention

[0005] The present invention provides a method and system for smart card chip defect detection based on deep learning, the main purpose of which is to improve the accuracy of existing chip defect detection.

[0006] To achieve the above objectives, the present invention provides a method for smart card chip defect detection based on deep learning, comprising: collecting thermoluminescence signals on the surface of the smart card chip using a thermoluminescence detector, and generating a defect thermal topology map of the surface of the smart card chip based on the thermoluminescence signals; Performing phonon inverse processing on the interior of the smart card chip according to the defect thermal topology map to obtain the phonon angular frequency inside the smart card chip; Based on the phonon angular frequency, constructing a phonon vibration frequency and virtual Joule heat distribution diagram of the smart card chip; injecting a modulated thermal wave into the smart card chip according to the phonon vibration frequency and the virtual Joule heat distribution diagram to collect in-phase thermal wave data and anti-phase thermal wave data of the smart card chip in response to the modulated thermal wave; The defective area detection result of the smart card chip is identified by the in-phase thermal wave data and the anti-phase thermal wave data.

[0007] Optionally, generating a defect thermal topology map of the surface of the smart card chip using the thermoluminescence signal includes: extracting a thermoluminescence intensity gradient, a thermoluminescence symmetry feature, and a thermoluminescence continuity feature from the thermoluminescence signal; When the thermoluminescence intensity gradient is greater than a preset intensity drop threshold, a region corresponding to the thermoluminescence intensity gradient on the surface of the smart card chip is regarded as a cavity defect region; When the intensity of the thermoluminescent symmetric feature is greater than a preset intensity deviation threshold, the area corresponding to the thermoluminescent symmetric feature on the surface of the smart card chip is used as a metal migration area; When the thermoluminescence continuous feature is greater than a preset mutation spacing threshold, a region corresponding to the thermoluminescence continuous feature on the surface of the smart card chip is used as a lattice distortion region; A defect thermal topology map of the smart card chip surface is generated based on the void defect area, the metal migration area, and the lattice distortion area.

[0008] Optionally, performing phonon inverse processing on the interior of the smart card chip according to the defect thermal topology map to obtain the phonon angular frequency inside the smart card chip includes: extracting the thermal conductivity of the smart card chip from the defect thermal topology map; calculating the interatomic force constant of the smart card chip according to the thermal conductivity; Performing phonon inverse processing on the interior of the smart card chip according to the interatomic force constant to obtain an initial phonon angular frequency inside the smart card chip; The initial phonon angular frequency is calibrated using a convolutional neural network to obtain the internal phonon angular frequency of the smart card chip.

[0009] Optionally, extracting the thermal conductivity of the smart card chip from the defect thermal topology map includes: Calculating a first thermal conductivity of a void defect region in the defect thermal topology map; Calculating a second differential value of the thermoluminescence signal intensity of the metal migration zone in the defect thermal topology map; determining a second thermal conductivity of the metal migration zone using the second differential value; calculating a third thermal conductivity of a lattice distortion region in the defect thermal topology map; The first thermal conductivity, the second thermal conductivity, and the third thermal conductivity are used as thermal conductivities of the smart card chip.

[0010] Optionally, constructing the phonon vibration frequency and virtual Joule heat distribution diagram of the smart card chip based on the phonon angular frequency includes: Converting the phonon angular frequency into the phonon vibration frequency of the smart card chip; Establishing a thermal-electric coupling model of the smart card chip according to the thermal conductivity corresponding to the phonon angular frequency, the operating voltage parameters and the electrical conductivity of the smart card chip; A virtual Joule heat distribution diagram of the smart card chip is determined by the thermal-electric coupling model.

[0011] Optionally, determining the virtual Joule heat distribution map of the smart card chip by using the thermal-electric coupling model includes: Determining the Joule heat power density of the smart card chip by using the thermal-electric coupling model; A virtual Joule heat distribution map of the smart card chip is constructed by taking the two-dimensional spatial coordinates of the smart card chip surface as the horizontal and vertical coordinates and taking the Joule heat power density as the hot spot intensity on the horizontal and vertical coordinates.

[0012] Optionally, injecting a modulated heat wave into the smart card chip according to the phonon vibration frequency and the virtual Joule heat distribution diagram includes: Obtaining a defect thermal topology map corresponding to the smart card chip; Inputting the phonon vibration frequency and the horizontal and vertical coordinates in the virtual Joule heat distribution map into a preset thermal actuator; injecting an in-phase modulated thermal wave into a defect region of the defect thermal topology map through the thermal actuator according to the phonon vibration frequency and the horizontal and vertical coordinates in the virtual Joule heat distribution map; On the basis of the in-phase modulated thermal wave, an anti-phase modulated thermal wave is injected into the non-defective area of ​​the defect thermal topology map to complete the process of injecting the modulated thermal wave into the smart card chip.

[0013] Optionally, collecting the in-phase thermal wave data and the anti-phase thermal wave data of the smart card chip in response to the modulated thermal wave includes: Detect phonon amplification in defective areas; Obtaining the hot spot offset belonging to the defect area in the virtual Joule heat distribution map; Using the phonon amplification and the hot spot offset as in-phase thermal wave data; Detect residual signals in non-defective areas; Obtaining the intensity of a second hot spot belonging to a non-defective area in the virtual Joule heat distribution map; The residual signal and the second hot spot intensity are used as anti-phase thermal wave data.

[0014] Optionally, identifying the defective area detection result of the smart card chip by using the in-phase thermal wave data and the anti-phase thermal wave data includes: Identifying a real defect area among defect areas by using a first comparison relationship between the in-phase thermal wave data and a preset first threshold; Identifying a false non-defective area in the non-defective area by using a second comparison relationship between the anti-phase thermal wave data and a preset second threshold; The real defect area and the false non-defect area are used as defect area detection results of the smart card chip.

[0015] In order to solve the above problems, the present invention also provides a smart card chip defect detection system based on deep learning, the system comprising: a defect generating module, configured to collect thermoluminescence signals on the surface of the smart card chip using a thermoluminescence detector, and generate a defect thermal topology map on the surface of the smart card chip based on the thermoluminescence signals; a phonon inversion module, configured to perform phonon inversion processing on the interior of the smart card chip according to the defect thermal topology map to obtain the phonon angular frequency inside the smart card chip; A distribution construction module, configured to construct a phonon vibration frequency and a virtual Joule heat distribution diagram of the smart card chip based on the phonon angular frequency; a data acquisition module, configured to inject a modulated thermal wave into the smart card chip according to the phonon vibration frequency and the virtual Joule heat distribution diagram, so as to acquire in-phase thermal wave data and anti-phase thermal wave data of the smart card chip in response to the modulated thermal wave; The defect detection module is configured to identify a defect area detection result of the smart card chip through the in-phase thermal wave data and the anti-phase thermal wave data.

[0016] Compared to the problems described in the background art, the present invention utilizes a thermoluminescence detector to collect thermoluminescence signals from the surface of the smart card chip, reducing the need for manual intervention and electrical signal testing in defect detection. The thermoluminescence signal can penetrate the chip packaging material, enabling micron-level defect location without damaging the chip structure. Based on the defect thermal topology map, the present invention performs phonon inversion processing on the interior of the smart card chip to analyze and obtain multi-source data, replacing traditional electromagnetic radiation or power signal data. This paves the way for subsequent multi-source data-based defect detection. The macroscopic thermal conductivity is converted into atomic-level phonon vibrations. Based on the phonon vibration frequency and the virtual Joule heat distribution map, the present invention injects modulated thermal waves into the smart card chip to induce a phonon resonance effect, increasing the defect signal amplification. The present invention uses the in-phase and anti-phase thermal wave data to identify the defect area detection results of the smart card chip, thereby determining the defect based on the physical signal of the chip response. Therefore, the deep learning-based smart card chip defect detection method and system provided by the present invention can improve the accuracy of existing chip defect detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A flowchart of a method for implementing smart card chip defect detection based on deep learning according to an embodiment of the present invention; Figure 2 A schematic diagram of a flow chart of modulated thermal wave injection for a smart card chip defect detection method based on deep learning provided by one embodiment of the present invention; Figure 3 A schematic diagram of a module for implementing a deep learning-based smart card chip defect detection system according to an 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] The present embodiment provides a method for detecting smart card chips based on deep learning. The method can be performed by at least one of the following electronic devices, including a server and a terminal, that can be configured to perform the method provided by the present embodiment. In other words, the method can be performed by software or hardware installed on a terminal or server. 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 FIG2 is a flow chart of a method for implementing a smart card chip defect detection method based on deep learning according to an embodiment of the present invention. In this embodiment, the method for implementing a smart card chip defect detection method based on deep learning includes: S1. Using a thermoluminescence detector to collect thermoluminescence signals on the surface of the smart card chip, and generating a defect thermal topology map of the surface of the smart card chip based on the thermoluminescence signals.

[0022] The embodiment of the present invention uses a thermoluminescence detector to collect thermoluminescence signals on the surface of the smart card chip to reduce manual participation and defect detection methods involving electrical signal testing. The thermoluminescence signal can penetrate the chip packaging material and achieve micron-level defect positioning without destroying the chip structure.

[0023] In the embodiment of the present invention, the thermoluminescence detector refers to a device that can detect and measure thermoluminescence phenomena. It is usually made of thermoluminescent materials (such as lithium fluoride, calcium sulfate, etc.). These materials can store energy when exposed to radiation and release photons during subsequent heating. The smart card chip surface refers to the outer surface of the smart card chip, that is, the part of the chip exposed outside the packaging material or the surface of the packaging material. In actual applications, smart card chips are usually packaged in materials such as plastic, but there may be defects on the surface of the packaging material, such as voids, metal migration, lattice distortion, etc. These defects will affect The generation and propagation of thermoluminescent signals. The thermoluminescent signal refers to the information carried by photons released from the thermoluminescent material. The intensity and number of these photons are related to the radiation dose received by the material. When the chip is heated by the heating device in the thermoluminescent detector, the chip releases thermal radiation. The thermoluminescent material inside the thermoluminescent detector absorbs the thermal radiation and stores energy. This energy is released in the form of photons. The resulting thermoluminescent signal is collected by the signal acquisition device in the thermoluminescent detector. The signal intensity can be measured by devices such as photomultiplier tubes and semiconductor photodetectors.

[0024] In one embodiment of the present invention, generating a defect thermal topology map of the smart card chip surface using the thermoluminescence signal includes: extracting a thermoluminescence intensity gradient, a thermoluminescence symmetry feature, and a thermoluminescence continuity feature from the thermoluminescence signal; when the thermoluminescence intensity gradient is greater than a preset intensity drop threshold, treating the area corresponding to the thermoluminescence intensity gradient on the smart card chip surface as a void defect area; when the thermoluminescence symmetry feature is greater than a preset intensity offset threshold, treating the area corresponding to the thermoluminescence symmetry feature on the smart card chip surface as a metal migration area; when the thermoluminescence continuity feature is greater than a preset mutation spacing threshold, treating the area corresponding to the thermoluminescence continuity feature on the smart card chip surface as a lattice distortion area; and generating a defect thermal topology map of the smart card chip surface based on the void defect area, the metal migration area, and the lattice distortion area.

[0025] The thermoluminescence intensity gradient refers to the spatial rate of change of the thermoluminescence signal intensity on the chip surface, reflecting the continuity of heat transfer. The formula for calculating the thermoluminescence intensity gradient uses the unit as the denominator and the numerator as the thermoluminescence intensity difference. When there are voids in the material, the heat conduction path is interrupted, resulting in abnormal temperature distribution above the void area, forming a high-intensity gradient. For example, at the edge of the void, the thermoluminescence intensity drops sharply from 5000 RLU to 800 RLU. The distance is within 2 microns, so the gradient value is (5000 - 800) / 2μm = 2100RLU / μm, the thermoluminescence symmetry feature refers to the symmetry difference in the spatial distribution of the thermoluminescence signal, which is quantified by the ellipse fitting parameters. The migration of metal ions leads to uneven distribution of material conductivity, causing hot spot offset, thereby destroying the thermoluminescence symmetry. For example, if the thermoluminescence signal is detected to be elliptical in distribution, with a major axis of 10μm and a minor axis of 3μm, the major axis divided by the minor axis = 3.33, which exceeds the intensity offset threshold of 1.8, it means that the thermoluminescence signal is asymmetric in spatial distribution at this time, that is, the intensity deviation of the thermoluminescence signals on the left and right sides is large, which is determined to be a metal migration area. The thermoluminescence continuity feature refers to the frequency of spatial mutation of the thermoluminescence signal. The density of high-frequency mutation points is detected by wavelet transform. This is because lattice distortion leads to stress concentration, forming microcracks or dislocations, causing discontinuous jumps in the thermoluminescence signal. For example, a signal jump occurs every 3μm in the lattice distortion area (mutation amplitude > 40%), the mutation point density reached 0.33 points / μm, exceeding the mutation spacing threshold of 0.2 points / μm. The defect thermal topology map refers to information on whether there are defects at different coordinates on the surface of the smart card chip, including defect areas and non-defect areas, as well as various detection data within each defect area. For example, if the coordinates (a, b) are a void defect area, that is, there is a defect, the void defect area includes coordinates, thermoluminescence signal data, etc.

[0026] Exemplarily, when the thermoluminescence intensity gradient is greater than a preset intensity drop threshold, the area corresponding to the thermoluminescence intensity gradient on the surface of the smart card chip is used as a void defect area. For example, when the thermoluminescence intensity gradient is greater than an intensity drop threshold such as 1500 RLU / μm, it indicates that there are air gaps or unfilled areas inside the material, which block the heat conduction path, resulting in a void defect.

[0027] S2. Perform phonon inverse processing on the interior of the smart card chip according to the defect thermal topology map to obtain the phonon angular frequency inside the smart card chip.

[0028] The embodiment of the present invention performs phonon inverse processing on the interior of the smart card chip based on the defect thermal topology map to analyze and obtain multi-source data, namely thermal conductivity and phonon angular frequency, replacing traditional electromagnetic radiation or power signal data, paving the way for subsequent multi-source data joint decision-making defect detection, and at the same time converting macroscopic thermal conductivity into atomic-level phonon vibrations.

[0029] In one embodiment of the present invention, performing phonon inverse processing on the interior of the smart card chip based on the defect thermal topology map to obtain the phonon angular frequency inside the smart card chip includes: extracting the thermal conductivity of the smart card chip from the defect thermal topology map; and calculating the interatomic force constant of the smart card chip based on the thermal conductivity using the following formula:

[0030] in, represents the interatomic force constant, represents thermal conductivity, represents the specific heat capacity, represents density, represents the speed of sound; According to the interatomic force constant, the following formula is used to perform phonon inverse processing on the interior of the smart card chip to obtain the initial phonon angular frequency inside the smart card chip:

[0031] in, represents the initial phonon angular frequency, represents the interatomic force constant, represents the atomic mass, represents the phonon wave vector modulus, represents the lattice constant; The initial phonon angular frequency is calibrated using a convolutional neural network to obtain the internal phonon angular frequency of the smart card chip.

[0032] The phonon angular frequency refers to the eigenfrequency of the lattice vibration, in rad / s. For example, the angular frequency of a normal silicon chip is ≈10.3×10¹³rad / s, and the void defect area is reduced to 7.5×10¹³rad / s, =π / the distance between adjacent atoms in the defect area, It is a commonly used formula in the prior art.

[0033] Optionally, the initial phonon angular frequency is calibrated using a convolutional neural network to obtain the internal phonon angular frequency of the smart card chip as follows: the interatomic force constant E, lattice constant a, and wave vector k′ of the defect area are output to the convolutional neural network, and the output is the phonon angular frequency correction value Δ , frequency after calibration ′= +Δ , so that the feature peak positioning error is reduced. The structure of the convolutional neural network is 3 layers of convolution (kernel size 3×3) + fully connected layer.

[0034] In another embodiment of the present invention, extracting the thermal conductivity of the smart card chip from the defect thermal topology map includes: calculating a first thermal conductivity of a void defect region in the defect thermal topology map using the following formula:

[0035] in, represents the first thermal conductivity, represents the thermal conductivity of the non-defective area in the defect thermal topology map, represents the thermoluminescence signal intensity of the void defect area, Indicates the thermoluminescence signal intensity of the non-defective area in the defect thermal topography; Calculate the second differential value of the thermoluminescence signal intensity of the metal migration region in the defect thermal topology; determine the second thermal conductivity of the metal migration region using the second differential value; and calculate the third thermal conductivity of the lattice distortion region in the defect thermal topology using the following formula:

[0036] in, represents the third thermal conductivity, represents the attenuation coefficient, represents the characteristic length of the chip, represents the gradient calculation result of the thermoluminescence signal intensity; The first thermal conductivity, the second thermal conductivity, and the third thermal conductivity are used as thermal conductivities of the smart card chip.

[0037] in, is the value of the pre-calibrated chip, 、 is the value directly measured by the thermoluminescence detector, and the calculation formula of the second differential value is: , the unit is ,about The value 1.8 is obtained from experimental data. For example, after fitting 50 sets of void samples, the prediction error of the exponent 1.8 is <5%.

[0038] Optionally, the process of determining the second thermal conductivity of the metal migration zone by using the second differential value is: according to When metal dendrites form bulges, The principle of interface delamination forming depression, if ,but (Dendrite region), if , (Layered area), that is, to set a fixed value for the second thermal conductivity, where 、 is the second thermal conductivity under different conditions.

[0039] S3. Constructing a phonon vibration frequency and virtual Joule heat distribution diagram of the smart card chip based on the phonon angular frequency.

[0040] In one embodiment of the present invention, constructing the phonon vibration frequency and virtual Joule heat distribution map of the smart card chip based on the phonon angular frequency includes: converting the phonon angular frequency into the phonon vibration frequency of the smart card chip; establishing a thermal-electric coupling model of the smart card chip based on the thermal conductivity corresponding to the phonon angular frequency, the operating voltage parameters and the electrical conductivity of the smart card chip; and determining the virtual Joule heat distribution map of the smart card chip through the thermal-electric coupling model.

[0041] Wherein, the thermal-electric coupling model is:

[0042] in, Indicates conductivity. When the chip is actually working, Changes with temperature T. For example, the σ of a copper conductor decreases by 40% at a temperature of 400K. The right side of the formula represents the process of Joule heat generating power, and the left side of the formula represents the heat dissipation capacity of the chip. The calculation formula for the change of σ with temperature is: , is the adjustable coefficient, represents thermal conductivity, represents the temperature gradient, represents the potential calculated from the voltage difference, The formula fully describes the coupled process of Joule heat generation and heat diffusion, takes into account the influence of temperature on electrical conductivity and thermal conductivity, ensures the accuracy of the calculation results, can handle complex boundary conditions and initial conditions, simulates the actual physical process, can perform transient analysis, and capture the change of temperature field and Joule heat power density over time. If only the right side of the formula is used ,Although the instantaneous value of Joule heat power density can be obtained, factors such as heat diffusion, the ,effect of temperature on material parameters, and boundary conditions cannot be considered, resulting inaccurate and ,incomplete results.

[0043] Optionally, the process of converting the phonon angular frequency into the phonon vibration frequency of the smart card chip is: using the formula The phonon angular frequency is converted into the phonon vibration frequency. Further, the process of determining the virtual Joule heat distribution diagram of the smart card chip through the thermal-electric coupling model is as follows: of The product of θ and thickness is taken as the Joule heat power density.

[0044] The parameter example table of the smart card chip regarding the thermal-electric coupling model is as follows:

[0045] In another embodiment of the present invention, determining the virtual Joule thermal distribution map of the smart card chip through the thermal-electric coupling model includes: determining the Joule thermal power density of the smart card chip through the thermal-electric coupling model; using the two-dimensional spatial coordinates of the surface of the smart card chip as the horizontal and vertical coordinates, and using the Joule thermal power density as the hot spot intensity on the horizontal and vertical coordinates, to construct a virtual Joule thermal distribution map of the smart card chip.

[0046] The horizontal and vertical coordinates refer to the coordinates on the X and Y axes.

[0047] S4. Injecting a modulated thermal wave into the smart card chip according to the phonon vibration frequency and the virtual Joule heat distribution diagram to collect in-phase thermal wave data and anti-phase thermal wave data of the smart card chip in response to the modulated thermal wave.

[0048] In the embodiment of the present invention, a modulated heat wave is injected into the smart card chip according to the phonon vibration frequency and the virtual Joule heat distribution diagram to induce a phonon resonance effect, thereby increasing the defect signal amplification.

[0049] In one embodiment of the present invention, the injecting of modulated thermal waves into the smart card chip according to the phonon vibration frequency and the virtual Joule heat distribution map includes: obtaining a defect thermal topology map corresponding to the smart card chip; inputting the phonon vibration frequency and the horizontal and vertical coordinates in the virtual Joule heat distribution map into a preset thermal actuator; injecting in-phase modulated thermal waves into the defect area of ​​the defect thermal topology map through the thermal actuator according to the phonon vibration frequency and the horizontal and vertical coordinates in the virtual Joule heat distribution map; and injecting anti-phase modulated thermal waves into the non-defect area of ​​the defect thermal topology map on the basis of the in-phase modulated thermal waves, so as to complete the process of injecting modulated thermal waves into the smart card chip.

[0050] The thermal actuator is used to calculate the phonon vibration frequency and the horizontal and vertical coordinates in the virtual Joule heat distribution diagram. The specific calculation formula is:

[0051] in, The spatial-temporal distribution function of thermal excitation comprehensively describes the intensity of thermal excitation at different locations and times on the chip surface (i.e., in-phase modulated thermal waves). represents the time-varying thermal excitation intensity, represents the maximum intensity of thermal excitation, represents the phonon vibration frequency, represents the coordinates of the thermal excitation center position, Indicates the coordinate position of the chip surface, represents the spatial distribution function of thermal excitation, describing the intensity distribution of thermal excitation at different positions on the chip surface, while the inverse phase modulated thermal wave is , which means the phase is shifted by 180 degrees.

[0052] See Figure 2 FIG. 1 is a flow chart of a modulated thermal wave injection method for implementing a smart card chip defect detection method based on deep learning according to an embodiment of the present invention. Figure 2 In the above, the defect prediction area refers to the defect area of ​​the defect thermal topology map, and the suspicious noise area refers to the non-defect area of ​​the defect thermal topology map.

[0053] In one embodiment of the present invention, the collection of the in-phase thermal wave data and the anti-phase thermal wave data of the smart card chip in response to the modulated thermal wave includes: detecting the phonon amplification of the defective area; obtaining the hot spot offset belonging to the defective area in the virtual Joule thermal distribution map; using the phonon amplification and the hot spot offset as the in-phase thermal wave data; detecting the residual signal of the non-defective area; obtaining the second hot spot intensity belonging to the non-defective area in the virtual Joule thermal distribution map; and using the residual signal and the second hot spot intensity as the anti-phase thermal wave data.

[0054] The phonon amplification is (the phonon energy when the thermal wave is on - the phonon energy when the thermal wave is off) / the phonon energy when the thermal wave is off 100%. For example, after a 12.3THz thermal wave is injected into a certain cavity defect area, the phonon energy increases from 80μJ to 108μJ, then the increase = 35%. The hot spot offset refers to the distance between the measured and predicted hot spot centers. The first hot spot intensity refers to the hot spot power density in the defect area. The residual signal is the residual phonon energy after the antiphase thermal wave / the phonon energy without excitation. For example, the residual energy after the antiphase thermal wave = 5.3μJ (basic energy = 100μJ), then the residual signal = 5.3%. The second hot spot intensity refers to the hot spot power density in the non-defect area.

[0055] S5. Identify a defective area detection result of the smart card chip using the in-phase thermal wave data and the anti-phase thermal wave data.

[0056] The embodiment of the present invention identifies the defect area detection result of the smart card chip through the in-phase thermal wave data and the anti-phase thermal wave data, so as to realize defect determination through the physical signal responded by the chip.

[0057] In one embodiment of the present invention, the identifying the defect area detection result of the smart card chip by the in-phase thermal wave data and the anti-phase thermal wave data includes: using a first comparison relationship between the in-phase thermal wave data and a preset first threshold to identify a real defect area in the defect area; using a second comparison relationship between the anti-phase thermal wave data and a preset second threshold to identify a false non-defect area in the non-defect area; and using the real defect area and the false non-defect area as the defect area detection result of the smart card chip.

[0058] Exemplarily, the first comparison relationship between the in-phase thermal wave data and the preset first threshold is used to identify the real defect area in the defect area. For example, if the hot spot offset is less than 1 μm and the phonon amplification is greater than 40%, it is determined to be a real defect area. Further, the second comparison relationship between the anti-phase thermal wave data and the preset second threshold is used to identify the false non-defect area in the non-defect area. For example, if the residual signal is greater than 5% or the second hot spot intensity is greater than 0.1 , it is determined that there is a defect, which is a false "non-defective area", that is, a true defective area.

[0059] Compared to the problems described in the background art, the present invention utilizes a thermoluminescence detector to collect thermoluminescence signals from the surface of the smart card chip, reducing the need for manual intervention and electrical signal testing in defect detection. The thermoluminescence signal can penetrate the chip packaging material, enabling micron-level defect location without damaging the chip structure. Based on the defect thermal topology map, the present invention performs phonon inversion processing on the interior of the smart card chip to analyze and obtain multi-source data, replacing traditional electromagnetic radiation or power signal data. This paves the way for subsequent multi-source data-based defect detection. The macroscopic thermal conductivity is converted into atomic-level phonon vibrations. Based on the phonon vibration frequency and the virtual Joule heat distribution map, the present invention injects modulated thermal waves into the smart card chip to induce a phonon resonance effect, increasing the defect signal amplification. The present invention uses the in-phase and anti-phase thermal wave data to identify the defect area detection results of the smart card chip, thereby determining the defect based on the physical signal of the chip response. Therefore, the deep learning-based smart card chip defect detection method and system provided by the present invention can improve the accuracy of existing chip defect detection.

[0060] Example 2: like Figure 3As shown in the figure, it is a functional module diagram of a defect detection system for smart card chips based on deep learning according to the present invention.

[0061] The deep learning-based smart card chip defect detection system 300 described in the present invention can be installed in an electronic device. Depending on the functionality implemented, the deep learning-based smart card chip defect detection system can include a defect generation module 301, a phonon inversion module 302, a distribution construction module 303, a data acquisition module 304, and a defect detection module 305. 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. These are stored in the electronic device's memory.

[0062] In the embodiment of the present invention, the functions of each module / unit are as follows: The defect generating module 301 is configured to collect thermoluminescence signals on the surface of the smart card chip using a thermoluminescence detector, and generate a defect thermal topology map on the surface of the smart card chip based on the thermoluminescence signals; The phonon inversion module 302 is configured to perform phonon inversion processing on the interior of the smart card chip according to the defect thermal topology map to obtain the phonon angular frequency inside the smart card chip; The distribution construction module 303 is used to construct a phonon vibration frequency and virtual Joule heat distribution diagram of the smart card chip based on the phonon angular frequency; The data acquisition module 304 is configured to inject a modulated thermal wave into the smart card chip according to the phonon vibration frequency and the virtual Joule heat distribution diagram, so as to acquire in-phase thermal wave data and anti-phase thermal wave data of the smart card chip in response to the modulated thermal wave; The defect detection module 305 is configured to identify a defect area detection result of the smart card chip through the in-phase thermal wave data and the anti-phase thermal wave data.

[0063] In detail, the modules in the smart card chip defect detection system 300 based on deep learning in the embodiment of the present invention are used in the same manner as above. Figure 1 The same technical means as the method for implementing smart card chip defect detection based on deep learning described in the previous section can produce the same technical effects, so I will not go into details here.

[0064] 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.

[0065] 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 smart card chip defect detection based on deep learning, characterized in that: The method comprises: collecting thermoluminescence signals on the surface of the smart card chip using a thermoluminescence detector, and generating a defect thermal topology map of the surface of the smart card chip based on the thermoluminescence signals; Performing phonon inverse processing on the interior of the smart card chip according to the defect thermal topology map to obtain the phonon angular frequency inside the smart card chip; Based on the phonon angular frequency, constructing a phonon vibration frequency and virtual Joule heat distribution diagram of the smart card chip; injecting a modulated thermal wave into the smart card chip according to the phonon vibration frequency and the virtual Joule heat distribution diagram to collect in-phase thermal wave data and anti-phase thermal wave data of the smart card chip in response to the modulated thermal wave; The defective area detection result of the smart card chip is identified by the in-phase thermal wave data and the anti-phase thermal wave data.

2. The method for smart card chip defect detection based on deep learning according to claim 1, characterized in that: Generating a defect thermal topology map of the surface of the smart card chip by using the thermoluminescence signal includes: extracting a thermoluminescence intensity gradient, a thermoluminescence symmetry feature, and a thermoluminescence continuity feature from the thermoluminescence signal; When the thermoluminescence intensity gradient is greater than a preset intensity drop threshold, a region corresponding to the thermoluminescence intensity gradient on the surface of the smart card chip is regarded as a cavity defect region; When the intensity of the thermoluminescent symmetric feature is greater than a preset intensity offset threshold, the area corresponding to the thermoluminescent symmetric feature on the surface of the smart card chip is used as a metal migration area; When the thermoluminescence continuous feature is greater than a preset mutation spacing threshold, a region corresponding to the thermoluminescence continuous feature on the surface of the smart card chip is used as a lattice distortion region; A defect thermal topology map of the smart card chip surface is generated based on the void defect area, the metal migration area, and the lattice distortion area.

3. The method for smart card chip defect detection based on deep learning according to claim 1, characterized in that: The step of performing phonon inverse processing on the interior of the smart card chip according to the defect thermal topology map to obtain the phonon angular frequency inside the smart card chip includes: extracting the thermal conductivity of the smart card chip from the defect thermal topology map; calculating the interatomic force constant of the smart card chip according to the thermal conductivity; Performing phonon inverse processing on the interior of the smart card chip according to the interatomic force constant to obtain an initial phonon angular frequency inside the smart card chip; The initial phonon angular frequency is calibrated using a convolutional neural network to obtain the internal phonon angular frequency of the smart card chip.

4. The method for smart card chip defect detection based on deep learning according to claim 3, characterized in that: The step of extracting the thermal conductivity of the smart card chip from the defect thermal topology map includes: Calculating a first thermal conductivity of a void defect region in the defect thermal topology map; Calculating a second differential value of the thermoluminescence signal intensity of the metal migration zone in the defect thermal topology map; determining a second thermal conductivity of the metal migration zone using the second differential value; calculating a third thermal conductivity of a lattice distortion region in the defect thermal topology map; The first thermal conductivity, the second thermal conductivity, and the third thermal conductivity are used as thermal conductivities of the smart card chip.

5. The method for smart card chip defect detection based on deep learning according to claim 1, wherein: The constructing of the phonon vibration frequency and virtual Joule heat distribution diagram of the smart card chip based on the phonon angular frequency includes: Converting the phonon angular frequency into the phonon vibration frequency of the smart card chip; Establishing a thermal-electric coupling model of the smart card chip according to the thermal conductivity corresponding to the phonon angular frequency, the operating voltage parameters and the electrical conductivity of the smart card chip; A virtual Joule heat distribution diagram of the smart card chip is determined by the thermal-electric coupling model.

6. The method for smart card chip defect detection based on deep learning according to claim 5, characterized in that: Determining a virtual Joule heat distribution diagram of the smart card chip by using the thermal-electric coupling model includes: Determining the Joule heat power density of the smart card chip by using the thermal-electric coupling model; A virtual Joule heat distribution map of the smart card chip is constructed by taking the two-dimensional spatial coordinates of the smart card chip surface as the horizontal and vertical coordinates and taking the Joule heat power density as the hot spot intensity on the horizontal and vertical coordinates.

7. The method for smart card chip defect detection based on deep learning according to claim 1, wherein: The injecting a modulated heat wave into the smart card chip according to the phonon vibration frequency and the virtual Joule heat distribution diagram comprises: Obtaining a defect thermal topology map corresponding to the smart card chip; Inputting the phonon vibration frequency and the horizontal and vertical coordinates in the virtual Joule heat distribution map into a preset thermal actuator; injecting an in-phase modulated thermal wave into a defect region of the defect thermal topology map through the thermal actuator according to the phonon vibration frequency and the horizontal and vertical coordinates in the virtual Joule heat distribution map; On the basis of the in-phase modulated thermal wave, an anti-phase modulated thermal wave is injected into the non-defective area of ​​the defect thermal topology map to complete the process of injecting the modulated thermal wave into the smart card chip.

8. The method for smart card chip defect detection based on deep learning according to claim 1, wherein: The collecting of the in-phase thermal wave data and the anti-phase thermal wave data of the smart card chip in response to the modulated thermal wave comprises: Detect phonon amplification in defective areas; Obtaining the hot spot offset belonging to the defect area in the virtual Joule heat distribution map; Using the phonon amplification and the hot spot offset as in-phase thermal wave data; Detect residual signals in non-defective areas; Obtaining the intensity of a second hot spot belonging to a non-defective area in the virtual Joule heat distribution map; The residual signal and the second hot spot intensity are used as anti-phase thermal wave data.

9. The method for smart card chip defect detection based on deep learning according to claim 1, wherein: The identifying the defective area detection result of the smart card chip by using the in-phase thermal wave data and the anti-phase thermal wave data includes: Identifying a real defect area among defect areas by using a first comparison relationship between the in-phase thermal wave data and a preset first threshold; Identifying a false non-defective area in the non-defective area by using a second comparison relationship between the anti-phase thermal wave data and a preset second threshold; The real defect area and the false non-defect area are used as defect area detection results of the smart card chip.

10. A smart card chip defect detection system based on deep learning, characterized in that: The system comprises: a defect generating module, configured to collect thermoluminescence signals on the surface of the smart card chip using a thermoluminescence detector, and generate a defect thermal topology map on the surface of the smart card chip based on the thermoluminescence signals; a phonon inversion module, configured to perform phonon inversion processing on the interior of the smart card chip according to the defect thermal topology map to obtain the phonon angular frequency inside the smart card chip; A distribution construction module, configured to construct a phonon vibration frequency and a virtual Joule heat distribution diagram of the smart card chip based on the phonon angular frequency; a data acquisition module, configured to inject a modulated thermal wave into the smart card chip according to the phonon vibration frequency and the virtual Joule heat distribution diagram, so as to acquire in-phase thermal wave data and anti-phase thermal wave data of the smart card chip in response to the modulated thermal wave; The defect detection module is configured to identify a defect area detection result of the smart card chip through the in-phase thermal wave data and the anti-phase thermal wave data.

Images