Intelligent Defect Detection System and Self-Healing Method for RFID Tags on Flexible Substrates

Abstract

This invention discloses an intelligent defect detection system and self-healing method for RFID tags on flexible substrates, belonging to the field of IoT electronic device manufacturing technology. The system integrates a high-resolution image acquisition module, a multispectral sensor array, and a robotic arm motion platform to construct a three-dimensional dynamic scanning system, capturing the surface and internal structural features of the flexible substrate in real time. A deep learning-based defect recognition algorithm, combined with multi-scale convolutional neural networks and transfer learning techniques, achieves accurate classification of 12 types of defects, including microcracks, conductive layer fractures, and substrate deformation, with a detection accuracy of 99.2%. A dual-mode self-healing mechanism is proposed: for conductive layer defects, nano-silver conductive colloid is injected through microfluidic channels, and 3D structural reconstruction is achieved using a controllable temperature field; for substrate damage, a photoresponsive shape memory polymer patch is used, which achieves molecular-level bonding repair after activation by ultraviolet light. This solution improves detection efficiency by more than 5 times, and the RF performance of the tag recovers to 98.7% of its initial value after self-healing.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) electronic device manufacturing technology, and in particular to an intelligent defect detection system and self-repair method for RFID tags on flexible substrates. Background Technology

[0002] Flexible substrate RFID tags, as core data carriers of the Internet of Things (IoT), involve precision printing, lamination, and encapsulation processes in their manufacturing. Currently, the industry commonly uses polyimide (PI) or polyethylene terephthalate (PET) as the substrate, coupled with a silver paste / copper foil conductive layer. However, under bending, humid heat, or mechanical stress, these tags are prone to microcracks (≤50μm), conductive layer breakage (linewidth loss ≥20%), and substrate delamination (area ≥0.1mm). 2 Defects such as [missing information] are present. In existing technologies, defect detection mainly relies on manual visual inspection (missing rate >35%) or single-spectrum imaging (such as visible light), while repair methods mostly use conductive adhesive bonding (resistance increase rate >15%) or epoxy resin filling (elastic modulus mismatch >40%), which is difficult to meet the production requirements of high precision and high reliability.

[0003] Referring to the Chinese patent "An RFID Tag Defect Detection System Based on Infrared Thermal Imaging," this technology heats the tag surface with a thermal excitation source and uses an infrared camera to capture local temperature rise differences in the conductive layer to identify short-circuit or open-circuit defects. However, this method has significant limitations:

[0004] Limited defect types: It can only detect defects related to the thermal effect of conductive layer current (such as short-circuit point temperature rise ≥5℃), and is ineffective for non-heating defects (such as substrate deformation, microcracks);

[0005] Insufficient resolution: Infrared imaging spatial resolution ≤100μm, unable to identify micron-level damage (<50μm);

[0006] High false alarm rate: Ambient temperature fluctuations (±2℃) cause the signal-to-noise ratio (SNR) to drop below 8dB, resulting in a false alarm rate exceeding 25%.

[0007] Based on the existing technologies described above, the following core issues still need to be addressed in the manufacturing of flexible substrate RFID tags:

[0008] Blind spots in multi-dimensional defect detection: Existing methods cannot simultaneously capture surface morphology (such as wrinkles and cracks) and internal structural defects (such as delamination and bubbles), resulting in an overall detection accuracy of less than 90%.

[0009] Deterioration of repair performance: The mechanical / electrical properties of traditional repair materials are mismatched with those of the substrate, resulting in a decrease in the bending life of the repaired label to less than 10,000 cycles (original value ≥ 100,000 cycles), and a ≥ 15% reduction in RF reading distance;

[0010] Lack of system synergy: The detection, repair, and verification processes are fragmented, the processing time for a single label is greater than 3 minutes, and there is a lack of a closed-loop optimization mechanism for process parameters. Summary of the Invention

[0011] In view of the aforementioned existing problems, the present invention is proposed.

[0012] Therefore, this invention provides an intelligent defect detection system and self-healing method for RFID tags on flexible substrates, which solves the problems of blind spots in multi-dimensional defect detection, deterioration of repair performance, and lack of system synergy.

[0013] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0014] In a first aspect, the present invention provides an intelligent RFID tag defect detection system for flexible substrates, comprising:

[0015] The high-resolution image acquisition module is equipped with a ring light source and an optical microscope head assembly to acquire micron-level morphological features of the flexible substrate surface. It achieves precise imaging of microcracks, wrinkles, and surface fracture defects of the conductive layer by eliminating interference from curved surface reflections.

[0016] A multispectral sensor array, integrating near-infrared and terahertz band sensors, is used to penetrate the surface of flexible substrates and detect internal conductive layer fractures, delamination, and internal bubble defects. Multi-band data fusion enhances the detection capability of hidden defects.

[0017] The six-degree-of-freedom robotic arm motion platform, equipped with the high-resolution image acquisition module and multispectral sensor array, performs full-coverage detection of flexible substrates through a three-dimensional dynamic scanning path. Combined with a sub-pixel-level image stitching algorithm, it achieves high-precision defect localization of large-size substrates. The defect recognition algorithm based on deep learning, combined with multi-scale convolutional neural networks and transfer learning technology, achieves accurate classification of 12 types of defects, including microcracks, conductive layer fractures, and substrate deformation.

[0018] As a preferred embodiment of the intelligent defect detection system for RFID tags on flexible substrates according to the present invention, the defect identification algorithm includes the following steps:

[0019] A multi-scale residual convolutional neural network model is constructed. The input layer receives multimodal features from high-resolution images and multispectral data. The backbone network contains three parallel convolutional branches with kernel sizes of 3×3, 5×5, and 7×7, respectively. The receptive field of each branch is expanded through dilated convolution.

[0020] The feature fusion layer uses a channel attention mechanism to dynamically weight the feature maps output by each branch. During the transfer learning stage, the pre-trained VGG-16 model weights are loaded and the shallow network parameters are frozen.

[0021] The output layer calculates the probability distribution of 12 defect classes using the Softmax function, and its classification accuracy satisfies:

[0022]

[0023] Where I(·) is the indicator function, N is the total number of test samples, and y i and These are the actual labels and the predicted labels, respectively.

[0024] As a preferred embodiment of the intelligent defect detection system for RFID tags on flexible substrates described in this invention, the improvement in detection efficiency is manifested in:

[0025] For a flexible substrate with dimensions of 200×200mm, the full inspection time T total The time is ≤30s, which is 5.2 times faster than traditional manual inspection, and it supports the identification of defects as small as 0.1mm.

[0026] As a preferred embodiment of the self-healing method for RFID tags with defects in flexible substrates according to the present invention, it includes the following steps:

[0027] S1: Three-dimensional scanning is completed through a robotic arm motion platform to simultaneously acquire surface morphology and internal structure data;

[0028] S2: A multi-scale residual convolutional neural network model is used to classify defects and output detection results with a confidence level of ≥95%; based on the defect classification results, a dual-mode self-healing mechanism is used to perform self-healing of the conductive layer or the substrate.

[0029] S3: If the defect is in the conductive layer, start the microfluidic injection-sintering process and control the temperature error to ≤±2℃;

[0030] S4: If the damage is to the substrate, match a pre-made shape memory polymer from the patch library and perform UV-activated bonding;

[0031] S5: Upload repair data to the cloud quality database through the edge computing unit, update the quality profile, and provide feedback on process optimization parameters.

[0032] As a preferred embodiment of the self-healing method for RFID tags with defects in flexible substrates according to the present invention, the self-healing method for the conductive layer in step S2 includes:

[0033] Step 1: Locate the defect region using microfluidic channels, based on the conductive layer fracture length L and the defect area S. defect Calculate the amount of silver nanoparticles injected, and the injection volume should satisfy:

[0034] V = πr 2 L+δ·S defect

[0035] Where r is the radius of the micro-nozzle, and δ = 1.2 to 1.5 is the safety redundancy factor;

[0036] Step 2: Under nitrogen protection, the defect area is heated to 120–150°C using pulse heating to trigger the sintering of the silver nanoparticles, forming a continuous conductive path. Its resistance recovery rate η R satisfy:

[0037]

[0038] Among them, R repaired For the repaired resistor, R initial The initial resistance;

[0039] Step 3: Use a laser Doppler vibration meter to test the mechanical strength of the repaired area, ensuring that the vibration amplitude error is less than ±5%.

[0040] As a preferred embodiment of the self-healing method for RFID tags with defects in flexible substrates according to the present invention, the self-healing method for the substrate in step S2 includes:

[0041] Step 1: Use a femtosecond laser to cut a shape memory polymer patch, the thickness h of which matches the mechanical properties of the substrate, satisfying the following:

[0042]

[0043] Among them, E subastrate E represents the elastic modulus of the flexible substrate. patch For the elastic modulus of the shape memory polymer patch, t substrate The thickness of the substrate;

[0044] Step 2: Use an ultraviolet light source (wavelength 365nm, power density ≥50mW / cm²) 2 The azophenyl groups on the surface of the activated patch trigger molecular chain reconstruction, and the bonding time t satisfies:

[0045]

[0046] Where, A = 1.2 × 10 3 B = 2.8 × 10 3 λ is a material constant, I is light intensity, and λ is wavelength.

[0047] Step 3: Use an atomic force microscope to check the roughness Ra of the repair interface, and ensure that Ra ≤ 0.1 μm.

[0048] As a preferred embodiment of the self-healing method for RFID tags with defects in flexible substrates according to the present invention, wherein: the dual-mode self-healing mechanism in step S2 automatically switches according to the defect depth:

[0049] When the defect depth d≥0.3, the substrate self-healing is initiated; otherwise, the conductive layer self-healing is executed. After the repair is completed, the tag ID is read through near-field communication and compared with historical parameters in the cloud quality database to generate a verification report containing the repair location, material usage, and performance indicators.

[0050] To address defects in the conductive layer, nano-silver conductive colloid is injected through microfluidic channels, and 3D structural reconstruction is achieved using a controllable temperature field. To address substrate damage, photoresponsive shape memory polymer patches are used, which are activated by ultraviolet light to achieve molecular-level bonding repair.

[0051] As a preferred embodiment of the self-healing method for RFID tags with defects in flexible substrates described in this invention, the radio frequency performance after repair satisfies the following:

[0052] Tag reading distance recovery rate:

[0053]

[0054] Among them, D repaired To read the distance after repair, D initial The initial reading distance is given, and the resonant frequency offset Δf ≤ ±0.5MHz.

[0055] As a preferred embodiment of the self-healing method for RFID tags with defects in flexible substrates according to the present invention, wherein: the edge computing unit in step S5:

[0056] An integrated FPGA accelerator performs real-time Gaussian filtering and histogram equalization preprocessing on the acquired data, with processing latency satisfying the following:

[0057]

[0058] Where, N pixel C represents the number of pixels per frame. op =5 represents the single-pixel operand, f FPGA =500MHz is the clock frequency.

[0059] As a preferred embodiment of the self-healing method for RFID tags with defects in flexible substrates according to the present invention, the cloud-based quality database in step S5 performs the following functions:

[0060] The system stores the process parameters, test results, and repair records for each batch of RFID tags. Production parameters are optimized through multivariate regression analysis, and a mapping model is established between antenna impedance Z and printing pressure P and temperature T.

[0061] Z = β0 + β1P + β2T + β3PT + β4P 2 +β5T 2

[0062] Where, β i The regression coefficient is dynamically adjusted so that |Z-50Ω| ≤ 0.5Ω.

[0063] In a second aspect, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the self-healing method for RFID tag defects on flexible substrates as described in the first aspect of the present invention.

[0064] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the self-healing method for RFID tag defects on flexible substrates as described in the first aspect of the present invention.

[0065] The beneficial effects of this invention are:

[0066] This invention significantly improves the accuracy and reliability of defect detection for RFID tags on flexible substrates through the synergistic application of high-resolution optical imaging and multispectral fusion technology. Employing an optical microscopic imaging module with a resolution of 0.5 μm, combined with a multispectral sensor array in the near-infrared and terahertz bands, the system can simultaneously capture hidden defects such as surface microcracks, wrinkles, and internal conductive layer fractures and delamination, achieving a detection accuracy of 99.2%, a 14.2 percentage point improvement over traditional single visible light detection methods. By using a ring light source and polarization filtering technology, reflection interference from curved substrates is effectively eliminated, increasing the imaging signal-to-noise ratio to over 35 dB. The minimum identifiable defect size is reduced to 0.1 mm, solving the problems of high false detection rates and insufficient resolution caused by environmental temperature fluctuations in traditional infrared thermal imaging technology. In terms of repair performance, the dual-mode self-healing mechanism proposed in this invention exhibits significant advantages: For conductive layer fracture defects, precise injection of nano-silver colloid through microfluidic channels combined with pulse sintering restores the resistivity of the repaired area to 98.7% of its initial value, reducing the resistance increase by 17 times compared to traditional conductive adhesive repair; for substrate damage, a light-responsive shape memory polymer patch with dynamically matched elastic modulus is used, achieving molecular-level bonding under UV light activation. The surface roughness of the repair interface is controlled at Ra≤0.1μm, and the bending life reaches 120,000 cycles, exceeding the durability of the original substrate. The system integrates a six-degree-of-freedom robotic arm motion platform and an edge computing unit, achieving 5cm scanning path via a helical involute. 2With a detection coverage efficiency of / s, combined with FPGA acceleration processing, the full inspection time for a single tag is reduced to 30 seconds, a 5.2-fold improvement over traditional manual operation. The cloud-based quality database dynamically optimizes production process parameters through a multivariate regression model, reducing antenna impedance fluctuations from ±3Ω to ±0.5Ω and establishing a defect root cause analysis mechanism, significantly improving production yield to 99.5%. This technology drives a 62% reduction in the overall cost per tag, with annual production capacity exceeding 50 million units. It is not only applicable to RFID tag manufacturing but can also be extended to cutting-edge fields such as flexible sensors and electronic skin, providing a complete technology solution for the large-scale reliable manufacturing of IoT devices. Attached Figure Description

[0067] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0068] Figure 1 This is a schematic diagram of the intelligent RFID tag defect detection system for flexible substrates in Example 1;

[0069] Figure 2 This is a flowchart of the RFID tag defect self-healing method for flexible substrates in Example 2. Detailed Implementation

[0070] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0071] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0072] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0073] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides an intelligent detection system for RFID tag defects on flexible substrates, comprising:

[0074] The high-resolution image acquisition module is equipped with a ring light source and an optical microscope head assembly to acquire micron-level morphological features of the flexible substrate surface. It achieves precise imaging of microcracks, wrinkles, and surface fracture defects of the conductive layer by eliminating interference from curved surface reflections.

[0075] A multispectral sensor array, integrating near-infrared and terahertz band sensors, is used to penetrate the surface of flexible substrates and detect internal conductive layer fractures, delamination, and internal bubble defects. Multi-band data fusion enhances the detection capability of hidden defects.

[0076] The six-degree-of-freedom robotic arm motion platform, equipped with the high-resolution image acquisition module and multispectral sensor array, performs full-coverage detection of flexible substrates through a three-dimensional dynamic scanning path. Combined with a sub-pixel-level image stitching algorithm, it achieves high-precision defect localization of large-size substrates. The defect recognition algorithm based on deep learning, combined with multi-scale convolutional neural networks and transfer learning technology, achieves accurate classification of 12 types of defects, including microcracks, conductive layer fractures, and substrate deformation.

[0077] The defect identification algorithm includes the following steps:

[0078] A multi-scale residual convolutional neural network model is constructed. The input layer receives multimodal features from high-resolution images and multispectral data. The backbone network contains three parallel convolutional branches with kernel sizes of 3×3, 5×5, and 7×7, respectively. The receptive field of each branch is expanded through dilated convolution.

[0079] The feature fusion layer uses a channel attention mechanism to dynamically weight the feature maps output by each branch. During the transfer learning stage, the pre-trained VGG-16 model weights are loaded and the shallow network parameters are frozen.

[0080] The output layer calculates the probability distribution of 12 defect classes using the Softmax function, and its classification accuracy satisfies:

[0081]

[0082] Where I(·) is the indicator function, N is the total number of test samples, and y i and These are the actual labels and the predicted labels, respectively.

[0083] The improvement in detection efficiency is manifested in:

[0084] For a flexible substrate with dimensions of 200×200mm, the full inspection time T total The time is ≤30s, which is 5.2 times faster than traditional manual inspection, and it supports the identification of defects as small as 0.1mm.

[0085] The specific implementation of the intelligent defect detection system for RFID tags on flexible substrates according to the present invention is as follows:

[0086] Upon system startup, the six-degree-of-freedom robotic arm motion platform initializes its positioning according to a preset 3D scanning path. Its end effector is equipped with a high-resolution image acquisition module and a multispectral sensor array. The robotic arm first uses a laser rangefinder to rapidly contour the surface of the flexible substrate, generating a 3D spatial coordinate model of the substrate. Subsequently, the robotic arm moves along a spiral involute trajectory, covering the substrate surface in a progressively expanding circular path in the transverse plane, while dynamically adjusting its height in the vertical direction to maintain the optimal working distance between the image acquisition module and the substrate surface. During the movement, the high-resolution image acquisition module continuously captures images of the substrate surface at a rate of 200 frames per second. The ring light source employs a multi-angle incident design, using polarizing filters to eliminate reflected light spots from curved surfaces. Combined with a 5-megapixel industrial camera and a 10x optical zoom lens, it can clearly capture surface cracks and conductive layer edge burrs as narrow as 0.5 micrometers. Each frame is transmitted in real-time to the edge computing unit, where an adaptive Gaussian filtering algorithm eliminates motion blur, and sub-pixel-level stitching technology integrates local images into a complete substrate surface morphology map.

[0087] A multispectral sensor array works synchronously with a high-resolution image acquisition module. As the robotic arm moves to each scanning node, the near-infrared sensor emits an 850nm wavelength light source to penetrate the substrate surface and receives the reflected spectral data to detect delamination defects at the interface between the conductive layer and the substrate. The terahertz sensor emits 0.3THz electromagnetic waves in pulse form and identifies bubbles or foreign matter embedded within the substrate by analyzing the phase changes of the transmitted waves. The raw data from both types of sensors are preprocessed by an edge computing unit and then spatiotemporally aligned with the surface topography image to form a multimodal fusion dataset. This dataset is transmitted via gigabit Ethernet to a server deploying a multi-scale residual convolutional neural network. The network first performs channel normalization on the input data, then extracts features at different scales through parallel convolutional branches. 3×3 convolutional kernels focus on microscopic texture details, 5×5 convolutional kernels capture medium-scale structural features, and 7×7 convolutional kernels identify macroscopic deformation patterns. The feature maps output from each branch are dynamically weighted and fused using a channel attention mechanism. Finally, a fully connected layer outputs the probability distribution of 12 types of defects. A classification result is triggered when the confidence level of a certain type of defect exceeds 95%.

[0088] When a conductive layer defect is detected, the system automatically switches to repair mode. The microfluidic repair unit locates the target area based on the defect coordinates, and a pneumatic clamp fixes the substrate to a vacuum adsorption platform to eliminate vibration interference. A micro-injection pump calculates the amount of nano-silver colloid injected based on the defect length and area, and precisely fills the fracture gaps using a pulse jet method through a quartz nozzle with an inner diameter of 50 micrometers. After injection, an infrared heating module under nitrogen atmosphere heats the temperature to 130°C at a rate of 10°C / second and maintains it for 30 seconds to allow the nano-silver particles to complete sintering and crystallization, forming a continuous conductive path. The repaired area is scanned and verified by a laser Doppler vibration meter, and the mechanical strength is confirmed to meet the standard by analyzing the vibration spectrum. If damage to the substrate body is detected, a robotic arm picks up a pre-made shape memory polymer patch from the patch library, and a femtosecond laser cutting system trims the patch edges to an accuracy of ±5 micrometers in real time according to the damage contour. The ultraviolet curing module uses a wavelength of 365nm and 80mW / cm². 2 The patch surface was irradiated with light for 10 seconds, activating the azobenzene functional groups and initiating a molecular chain cross-linking reaction. Simultaneously, a robotic arm applied 5N of pressure to ensure full contact between the patch and the substrate interface. After repair, atomic force microscopy was used to scan the nanoscale morphology of the interface area, and surface roughness analysis was used to verify the bonding quality.

[0089] The edge computing unit monitors the status of each subsystem in real time throughout the process, dynamically optimizing image processing, motion control, and repair parameters. For example, when a significant change in the curvature of a batch of substrates is detected, the Z-axis motion compensation of the robotic arm is automatically adjusted; or the penetration depth model of the terahertz sensor is corrected based on changes in ambient temperature and humidity. All detection data and repair records are uploaded to the cloud-based quality database via a 5G module. The system trains a process optimization model based on historical data. When analysis reveals that the impedance deviation of a certain label model is mainly due to fluctuations in printing pressure, an adjustment command is automatically sent to the production line, improving the pressure control accuracy from ±2N to ±0.5N. Simultaneously, the cloud platform generates a visualized quality report, allowing traceability of the detection time, defect type, repair material usage, and performance verification data for each label, providing data support for the entire product lifecycle management.

[0090] Example 2, refer to Figure 2 This is a second embodiment of the present invention, which provides a self-healing method for RFID tags with defects in flexible substrates, including the following steps:

[0091] S1: Three-dimensional scanning is completed through a robotic arm motion platform to simultaneously acquire surface morphology and internal structure data;

[0092] In the specific implementation of step S1, the robotic arm motion platform first completes initial calibration through the substrate positioning module. After the operator lays the flexible substrate flat on the vacuum adsorption worktable, the system starts a self-test program: the laser ranging multispectral sensor array at the end of the robotic arm quickly scans along the four corners of the substrate, emitting a 650nm wavelength laser beam and receiving the reflected signal. It calculates the three-dimensional coordinates of each point on the substrate surface using the time-of-flight (ToF) principle to generate an initial spatial model. Subsequently, the path planning algorithm automatically generates a spiral involute scanning path based on the substrate size (e.g., 200mm × 200mm) and curvature distribution—the robotic arm starts from the center point of the substrate and expands outward along a clockwise spiral trajectory with an increasing radius. The radius increment for each revolution is set to 5mm. At the same time, the Z-axis servo motor dynamically adjusts the height according to the real-time ranging data, so that the image acquisition module and the substrate surface always maintain the optimal focusing distance of 2.5±0.1mm.

[0093] During the movement, the high-resolution image acquisition module and the multispectral sensor array operate in a synchronized trigger mode. When the robotic arm moves to each path node, the motion controller sends a synchronization pulse signal: first, it triggers the ring light source of the high-resolution camera, using 12 LED strips at a 45° angle to illuminate the substrate surface, combined with a linear polarization filter to eliminate specular reflection interference. The camera continuously captures three frames with an exposure time of 1 / 1000 second, and a multi-frame super-resolution algorithm is used to synthesize a clear image with a single resolution of 5μm / pixel, which is then transmitted in real-time to the edge computing unit for distortion correction and local image stitching. Meanwhile, the near-infrared module in the multispectral sensor array emits modulated light in three bands: 850nm, 1300nm, and 1550nm. The reflected light is guided to the InGaAs detector by a beam splitter, and a layered structure map of 0.1-0.5mm depth below the substrate surface is constructed based on the differences in absorption rate of each band. The terahertz module emits broadband pulse waves of 0.1-1THz, and the transmitted signal is received by a high-sensitivity Schottky diode. Three-dimensional volume data of 1-3mm depth inside the substrate is generated by time-domain spectral analysis.

[0094] All sensor data are embedded with high-precision timestamps (error ≤ 1μs) and spatial coordinate labels during acquisition. The edge computing unit maps image pixels and spectral voxels to a unified three-dimensional coordinate system using a spatiotemporal alignment algorithm. For example, when the robotic arm moves to coordinates (X = 75.3mm, Y = 43.8mm, Z = 2.6mm), the surface image block acquired at that location is labeled as the spatial grid G(75,43), and the corresponding near-infrared spectral data slice is associated with the Z-axis layers L1-L5. The data processing pipeline adopts a double-buffering mechanism: while the data of the current scan point is being extracted, the data of the next node has already been transferred to the DDR4 cache through the PCIe interface, ensuring uninterrupted scanning motion. After completing the full substrate scan, the system automatically generates a fused dataset—the surface morphology is presented in RGB point cloud form, and the internal structure is displayed using pseudo-color encoding to show layered defects and foreign object embeddings. All abnormal areas are labeled with three-dimensional bounding boxes for subsequent defect classification module calls.

[0095] Throughout the scanning process, the environmental monitoring subsystem continuously maintains a constant temperature (25±0.5℃) and humidity (40%±3%RH) in the working area, and eliminates mechanical vibration interference with frequencies >5Hz through an active vibration damping platform. When the detected substrate edge warping exceeds a preset threshold (e.g., Z-axis offset >0.8mm), the robotic arm automatically switches to adaptive scanning mode: superimposing a sinusoidal waveform compensation motion on the helical path to ensure that the sensor probe is always perpendicular to the substrate surface. After scanning, the system generates a detection quality report, including data integrity verification (e.g., coverage ≥99.8%), a signal-to-noise ratio distribution heatmap, and a preview of abnormal areas, providing operators with a visual interactive interface to confirm the validity of the detection.

[0096] S2: A multi-scale residual convolutional neural network model is used to classify defects and output detection results with a confidence level of ≥95%; based on the defect classification results, a dual-mode self-healing mechanism is used to perform self-healing of the conductive layer or the substrate.

[0097] The self-healing method for the conductive layer in step S2 includes:

[0098] Step 1: Locate the defect region using microfluidic channels, based on the conductive layer fracture length L and the defect area S. defect Calculate the amount of silver nanoparticles injected, and the injection volume should satisfy:

[0099] V = πr 2 L+δ·S defect

[0100] Where r is the radius of the micro-nozzle, and δ = 1.2 to 1.5 is the safety redundancy factor;

[0101] Step 2: Under nitrogen protection, the defect area is heated to 120–150°C using pulse heating to trigger the sintering of the silver nanoparticles, forming a continuous conductive path. Its resistance recovery rate η R satisfy:

[0102]

[0103] Among them, R repaired For the repaired resistor, R initial The initial resistance;

[0104] Step 3: Use a laser Doppler vibration meter to test the mechanical strength of the repaired area, ensuring that the vibration amplitude error is less than ±5%.

[0105] The substrate self-healing method in step S2 includes:

[0106] Step 1: Use a femtosecond laser to cut a shape memory polymer patch, the thickness h of which matches the mechanical properties of the substrate, satisfying the following:

[0107]

[0108] Among them, E substrate E represents the elastic modulus of the flexible substrate. patch For the elastic modulus of the shape memory polymer patch, t substrate The thickness of the substrate;

[0109] Step 2: Use an ultraviolet light source (wavelength 365nm, power density ≥50mW / cm²) 2 The azophenyl groups on the surface of the activated patch trigger molecular chain reconstruction, and the bonding time t satisfies:

[0110]

[0111] Where, A = 1.2 × 10 3 B = 2.8 × 10 3 λ is a material constant, I is light intensity, and λ is wavelength.

[0112] Step 3: Use an atomic force microscope to check the roughness Ra of the repair interface, and ensure that Ra ≤ 0.1 μm.

[0113] The dual-mode self-healing mechanism in step S2 automatically switches according to the defect depth:

[0114] When the defect depth d≥0.3, the substrate self-healing is initiated; otherwise, the conductive layer self-healing is executed. After the repair is completed, the tag ID is read through near-field communication and compared with historical parameters in the cloud quality database to generate a verification report containing the repair location, material usage, and performance indicators.

[0115] To address defects in the conductive layer, nano-silver conductive colloid is injected through microfluidic channels, and 3D structural reconstruction is achieved using a controllable temperature field. To address substrate damage, photoresponsive shape memory polymer patches are used, which are activated by ultraviolet light to achieve molecular-level bonding repair.

[0116] The radio frequency performance after repair meets the following requirements:

[0117] Tag reading distance recovery rate:

[0118]

[0119] Among them, D repaired To read the distance after repair, D initial The initial reading distance is given, and the resonant frequency offset Δf ≤ ±0.5MHz.

[0120] In the specific implementation of step S2, the defect classification process of the multi-scale residual convolutional neural network model is realized through a precisely designed computational architecture and data processing pipeline. The system first performs standardized preprocessing on the multimodal fusion dataset obtained in step S1: the high-resolution surface image is normalized to map pixel values ​​to the range of 0 to 1; the near-infrared and terahertz spectral data are unified in dimensions using the min-max scaling method and spatially aligned with the image data to form an input tensor of dimension 512×512×5 (containing five feature layers: RGB three-channel surface image, near-infrared reflectance, and terahertz transmittance). The data loader inputs the preprocessed data to the GPU-accelerated computing nodes in batch processing.

[0121] The network model employs a three-branch parallel structure, with each branch corresponding to feature extraction at different scales: the first branch uses 3×3 convolutional kernels with dilated convolutional layers of 2 to capture micron-level texture features (such as crack orientation and irregularities at the edges of conductive layers); the second branch uses 5×5 convolutional kernels with max-pooling layers to extract medium-scale structural anomalies (such as local wrinkles and bubble outlines); the third branch uses 7×7 convolutional kernels combined with spatial pyramid pooling to identify macroscopic deformation patterns (such as overall warping and large-area layering). Each convolutional layer is followed by batch normalization and ReLU activation functions, and shallow and deep features are fused through skip connections to avoid the gradient vanishing problem.

[0122] The feature fusion stage introduces a channel attention mechanism: the feature map output from each branch first undergoes global average pooling to generate channel description vectors, and then the weight coefficients of each channel are learned through a two-layer fully connected network. Subsequently, the original feature maps are channel-weighted. The three weighted feature maps are then concatenated along the channel dimension to form a fused feature containing multi-scale information. The transfer learning module loads the weights of a VGG-16 model pre-trained on the ImageNet dataset, freezes the parameters of the first three convolutional blocks, and only fine-tunes the high-level networks, accelerating model convergence and improving generalization ability in small-sample scenarios.

[0123] In the classification decision stage, the fully connected layer maps the fused features to a 12-dimensional output vector, corresponding to predefined defect categories such as microcracks, conductive layer fractures, and substrate delamination. The Softmax function normalizes the output vector, generating probability distributions for each category. When the predicted probability of a certain category exceeds a 95% threshold, the system determines that the defect type is valid and marks the result as a high-confidence detection. If the highest probability is below 95%, a verification mechanism is triggered, extracting region proposals from the feature map, performing local magnification and secondary feature extraction on the suspected region, until the confidence requirement is met or the maximum number of iterations (default 3) is reached.

[0124] During model training, an adaptive momentum optimizer (Adam) was used with an initial learning rate of 0.001, which decayed by 50% every 20 epochs. The loss function employed weighted cross-entropy, with a 3x weighting coefficient applied to minority class defects (such as internal bubbles in the substrate) to balance the sample distribution. Training data augmentation strategies included random rotation (±5°), Gaussian noise injection (σ = 0.01), and elastic deformation (α = 30) to improve the model's robustness to disturbances such as light fluctuations and mechanical vibrations in real-world production environments. If the validation set accuracy did not improve for 10 consecutive epochs, an early stopping mechanism was automatically initiated to save the optimal model parameters.

[0125] During inference, the edge computing unit quantizes and compresses the model using the TensorRT engine, converting floating-point calculations to INT8 precision. This reduces the single-frame inference time to less than 8ms while ensuring a classification accuracy loss of less than 0.5%. The classification results are bound to the original data spatial coordinates, generating a structured report containing defect type, confidence level, 3D location, and bounding box size, which is pushed to the repair control subsystem in real time via a message queue. Simultaneously, the system continuously monitors model performance metrics. When the false alarm rate for a certain type of defect continuously exceeds 0.3%, an online learning process is automatically triggered, updating the training set with newly collected samples and initiating model fine-tuning to achieve dynamic evolution of defect classification capabilities.

[0126] S3: If the defect is in the conductive layer, start the microfluidic injection-sintering process and control the temperature error to ≤±2℃;

[0127] In the specific implementation of step S3, once the system determines the existence of a conductive layer defect and completes precise positioning, the microfluidic repair unit immediately initiates a fully automated processing procedure. The robotic arm first transfers the vacuum adsorption platform carrying the defective substrate to a sealed nitrogen-protected chamber. An oxygen concentration sensor inside the chamber monitors and maintains the concentration below 50 ppm in real time to prevent oxidation of the nano-silver colloid at high temperatures. The repair probe assembly moves above the defect coordinates via a precision slide rail. Its end integrates a three-axis micro-motion platform (positioning accuracy ±1 μm), carrying a quartz micro-nozzle with an inner diameter of 50 μm and an infrared temperature measurement module. The visual-assisted positioning system performs a secondary verification of the defect area using a coaxial optical microscope camera. Based on an image edge detection algorithm, it confirms the precise direction and width of the fracture crack and generates a three-dimensional path plan. If it is a straight fracture, continuous injection is performed along the crack centerline at a speed of 0.2 mm / s; if it is a network crack, a reciprocating filling strategy is adopted to ensure that the colloid covers all branch cracks.

[0128] The microfluidic control system dynamically adjusts the injection parameters based on the defect geometry: the injection pump, driven by piezoelectric ceramics, responds to flow control commands within 0.1 ms, injecting pre-loaded nano-silver colloid (30 nm particle size, 85% solid content) into the crack in a pulse jet mode. Each pulse cycle includes a 20 ms pressurized injection (flow rate 0.5 μL / s) and a 10 ms holding pressure phase to prevent colloid backflow. During injection, a near-infrared thermal imager monitors the temperature distribution at the colloid flow front at a rate of 100 frames per second. When an abnormal local temperature increase (>3℃) is detected, the injection is automatically paused and an ultrasonic vibration module (frequency 40 kHz, amplitude 5 μm) is activated to eliminate microbubbles or impurities. After injection, the system initiates a multi-stage drying procedure: first, purging with room temperature nitrogen for 10 seconds to remove volatile solvents, then pre-curing the colloid through gradient heating (50℃ / 2 min → 80℃ / 1 min).

[0129] The sintering stage employs a zoned temperature control strategy, with a 2mm diameter focused infrared heating ring positioned above the defect area. This ring integrates a platinum resistance temperature sensor (accuracy ±0.1℃) and a dual-feedback system of a spectroradiometer. The temperature control module executes the sintering process according to a preset curve: in the initial stage, the temperature is increased to 120℃ at a rate of 10℃ / s and maintained for 30 seconds to allow for initial densification of the colloid; in the second stage, the temperature is increased to 150℃ at a rate of 5℃ / s and maintained for 20 seconds to complete the neck fusion of the silver nanoparticles. Throughout the process, the PID controller collects the actual temperature every 10ms and strictly controls temperature fluctuations within ±2℃ by adjusting the output power of the infrared laser (adjustment accuracy 0.1W). When local overheating is detected, a micro-vortex tube cooling device is immediately activated, injecting -10℃ nitrogen gas for dynamic cooling of the overheated area.

[0130] After sintering, a four-probe resistance meter automatically presses down onto the repair area, applies a constant current of 10mA to measure the voltage drop, and calculates the line resistance value of the repair path. Simultaneously, a laser Doppler vibrometer applies a 0-1kHz sweep frequency vibration excitation to the surface of the repair area, assessing structural integrity by analyzing the resonant frequency shift. If the resistance recovery rate does not reach 98% or abnormal peaks appear in the vibration spectrum, the system automatically marks the point for secondary repair. First, a plasma cleaning gun (argon flow rate 5L / min, power 100W) is used to remove the surface oxide layer, then the injection and sintering process is repeated. All process parameters (such as injection volume, temperature profile, and resistance measurement values) are recorded in the quality traceability system and linked to the electronic history of the RFID tag for subsequent product reliability analysis. The entire repair process is completed within 120 seconds, during which the environmental control system continuously maintains a slightly positive pressure (50Pa) inside the chamber to isolate external dust contamination.

[0131] S4: If the damage is to the substrate, match a pre-made shape memory polymer from the patch library and perform UV-activated bonding;

[0132] In the specific implementation of step S4, when the system determines that there is damage to the substrate, the repair process immediately switches to shape memory polymer patch bonding mode. The robotic arm first sends the 3D contour data of the damaged area to the patch library management system, which is based on the damaged area (e.g., 5mm). 2 Parameters such as depth (e.g., 0.3 mm) and contour complexity (e.g., edge curvature radius ≤ 0.1 mm) are used to screen candidate patches from a prefabricated multi-layer storage rack, selecting those with an elastic modulus matching degree ≥ 95%. The patch material is a polyurethane shape memory polymer doped with azophenyl groups, with a glass transition temperature set at 60℃, and is pre-stored in a sealed cartridge under constant temperature and humidity (25℃, 30% RH). A vacuum suction cup tool is fitted to the end effector of the robotic arm to extract the target patch from the cartridge with a suction force of 0.1N. During the transfer process, the flatness of the patch is monitored in real time using a laser rangefinder to ensure that there is no twisting or deformation during transportation.

[0133] Once the patch reaches the damaged area, the femtosecond laser cutting system initiates a refinement process: the coaxial vision system captures the damaged outline at 500x magnification, the image processing algorithm extracts edge feature points to generate a vector path, and the femtosecond laser (wavelength 1030nm, pulse width 300fs) is controlled to perform contour cutting at a frequency of 2000 times per second. The laser focus diameter is controlled at 10μm, and the cutting path maintains a 50μm interference fit with the damaged edge to ensure maximum contact area between the patch and the substrate. During the cutting process, the blowing nozzle continuously sprays cryogenic argon gas (-20℃) to prevent the heat-affected zone from expanding, achieving a cutting accuracy of ±5μm. The refined patch is transferred to the pre-alignment station, and the robotic arm switches to six-dimensional force control mode. Using an array of micro-force sensors (range 0~10N, resolution 0.01N), the contact pressure is sensed, and the patch is applied to the damaged area at a speed of 0.5mm / s. During this process, the pitch angle is adjusted in real time to ensure that the patch surface matches the curvature of the substrate.

[0134] The UV activation stage employs a dynamic exposure strategy: an LED array light source (wavelength 365nm±5nm) positioned at the top of the repair chamber provides gradient irradiation in three zones, with the light intensity in the central zone set to 80mW / cm². 2 The edge transition zone is 60mW / cm 2 The unirradiated area in the outer perimeter is 20mW / cm². 2 The light intensity distribution is spatially modulated using a digital micromirror device (DMD), generating a matching mask pattern in real time based on the patch shape. The activation process consists of two phases: continuous irradiation for the first 30 seconds raises the surface temperature of the patch to 45°C, triggering the glass transition; the subsequent 60 seconds switch to a 10Hz pulse mode (50% duty cycle), using thermal relaxation control to prevent localized overheating. During irradiation, a robotic arm maintains a vertical pressure of 5N and promotes interfacial molecular diffusion through micro-vibrations (20μm amplitude, 5Hz frequency).

[0135] After bonding is completed, the in-situ quality inspection module is immediately activated. A confocal white light interferometer performs a three-dimensional morphological scan of the repair interface, acquiring surface height data in 1μm steps along the X / Y directions. The spatial frequency components are analyzed using fast Fourier transform to calculate the surface roughness Ra value. If a local area with Ra > 0.1μm is detected, the system automatically marks the location and initiates a secondary pressing procedure: a micro hydraulic punch applies an instantaneous pressure of 20MPa at the target point, combined with local supplemental ultraviolet irradiation (100mW / cm²). 2Strengthening bonding is performed (within 10 seconds). The repaired area that passes acceptance will undergo a bending fatigue test. A robotic arm will hold both ends of the substrate and perform 100 reciprocating bends (radius of curvature 5mm, frequency 1Hz). During this time, strain sensors will monitor the stress distribution at the patch edges to ensure that the maximum strain value does not exceed 105% of the original substrate value. All process parameters (such as light dose, pressure profile, and roughness data) are integrated into the RFID tag's digital archive, providing a complete data chain for subsequent reliability assessments.

[0136] S5: Upload repair data to the cloud quality database through the edge computing unit, update the quality profile, and provide feedback on process optimization parameters.

[0137] Edge computing unit in step S5:

[0138] An integrated FPGA accelerator performs real-time Gaussian filtering and histogram equalization preprocessing on the acquired data, with processing latency satisfying the following:

[0139]

[0140] Where, N pixel C represents the number of pixels per frame. op =5 represents the single-pixel operand, f FPGA =500MHz is the clock frequency.

[0141] The cloud-based quality database in step S5 implements the following functions:

[0142] The system stores the process parameters, test results, and repair records for each batch of RFID tags. Production parameters are optimized through multivariate regression analysis, and a mapping model is established between antenna impedance Z and printing pressure P and temperature T.

[0143] Z = β0 + β1P + β2T + β3PT + β4P 2 +β5T 2

[0144] Where, β i The regression coefficient is dynamically adjusted so that |Z-50Ω| ≤ 0.5Ω.

[0145] In the specific implementation of step S5, the edge computing unit first performs structured encapsulation of the multidimensional data generated during the repair process. Once the self-repair process of the conductive layer or substrate is complete, the system automatically triggers a data acquisition thread: extracting high-resolution surface topography images (5 μm per pixel) and terahertz voxel data (0.1 mm resolution) from the multispectral sensor array. 3The data includes 18 categories of key parameters, such as process parameters (e.g., sintering temperature curve, UV exposure dose) and performance verification indicators (resistivity recovery rate, roughness value). The data cleaning module built into the edge node performs noise reduction on this raw information—using sliding window mean filtering to eliminate temperature acquisition noise, correcting time deviations of multi-source data through timestamp alignment algorithms, and converting unstructured text logs (e.g., device status codes) into standard fields in JSON format.

[0146] The cleaned data packets are transmitted through a dual-encryption channel: first, the payload is encrypted using the AES-256 algorithm, and then a secure connection to the cloud is established through an industrial-grade VPN tunnel. The transmission protocol adopts an optimized MQTT / SSL hybrid architecture, achieving an average transmission rate of 120Mbps in a 5G network environment. In the event of network interruption, a local SSD cache (1TB capacity) is automatically activated, supporting breakpoint resumption and data integrity verification (SHA-256 hash verification). Each data packet is appended with a metadata header containing the device ID, timestamp, and batch number, ensuring traceability of the data source in the cloud.

[0147] The cloud-based quality database adopts a distributed time-series database architecture, establishing a three-layer storage structure based on the unique tag ID: the raw layer stores binary sensor data, the parsing layer stores structured data after feature extraction (such as defect coordinates and repair area), and the application layer integrates process parameter optimization models. When new data arrives, the real-time computing engine first performs correlation analysis, calculating the Pearson correlation between the conductive layer resistance recovery rate of the current repair record and the printing pressure and curing temperature of historical tags of the same model. If a correlation is found between the pressure parameter P and the resistance recovery rate η... R If the correlation coefficient |r|>0.7, the parameter optimization process is triggered. The machine learning model (such as the XGBoost regressor) dynamically generates printing pressure adjustment suggestions based on the mapping relationship obtained by training on hundreds of thousands of historical samples. For example, it can fine-tune the current production line pressure setting from 23.5N to 24.2N and send the correction command to the workshop PLC controller through the OPC UA protocol.

[0148] Meanwhile, the digital twin subsystem constructs a virtual production line model in the cloud, mapping real-time uploaded inspection data and 3D repair records to corresponding positions on virtual labels. Quality engineers can perform virtual sectioning analysis via a web interface: selecting any label ID retrieves its complete manufacturing file, including infrared thermograms of the original defects, atomic force microscopy topological maps of the repaired area, and stress-strain curves from bending tests. The system automatically generates a quality trend dashboard that displays key indicators in real time, such as hourly statistics on substrate delamination defect rates and conductive layer self-repair success rates. When the increase in the same type of defect exceeds 0.5% for three consecutive batches, an alert is triggered, and optimization solutions are pushed to mobile terminals.

[0149] In the final stage of the feedback loop, the cloud compiles the condensed process knowledge (such as the optimized sintering temperature gradient function and patch thickness calculation formula) into a binary instruction set executable by the edge nodes, and incrementally deploys it to each testing station through a differential update mechanism. For example, when a new flexible substrate is put into production, the system automatically loads the matching scanning path parameter library and adjusts the radius increment of the robotic arm's helical trajectory from 5mm to 3mm to accommodate a smaller radius of curvature. All data interaction processes are recorded in the blockchain notarization module, forming an immutable quality traceability chain that meets the ISO 9001 standard's requirements for the full lifecycle management of electronic quality records.

[0150] This embodiment also provides a computer device applicable to the self-repair method for RFID tag defects on flexible substrates, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the self-repair method for RFID tag defects on flexible substrates as proposed in the above embodiment.

[0151] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0152] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the self-repair method for RFID tag defects on flexible substrates as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0153] In summary, this invention significantly improves the accuracy and reliability of defect detection for flexible substrate RFID tags through the synergistic application of high-resolution optical imaging and multispectral fusion technology. Employing an optical microscopic imaging module with a resolution of 0.5 μm, combined with a multispectral sensor array in the near-infrared and terahertz bands, the system can simultaneously capture hidden defects such as surface microcracks, wrinkles, and internal conductive layer fractures and delamination, achieving a detection accuracy of 99.2%, a 14.2 percentage point improvement over traditional single visible light detection methods. By using a ring light source and polarization filtering technology to effectively eliminate reflection interference from curved substrates, the imaging signal-to-noise ratio is improved to over 35 dB, and the minimum identifiable defect size is reduced to 0.1 mm, solving the problems of high false detection rates and insufficient resolution caused by environmental temperature fluctuations in traditional infrared thermal imaging technology. In terms of repair performance, the dual-mode self-healing mechanism proposed in this invention exhibits significant advantages: For conductive layer fracture defects, precise injection of nano-silver colloid through microfluidic channels combined with pulse sintering restores the resistivity of the repaired area to 98.7% of its initial value, reducing the resistance increase by 17 times compared to traditional conductive adhesive repair; for substrate damage, a light-responsive shape memory polymer patch with dynamically matched elastic modulus is used, achieving molecular-level bonding under UV light activation. The surface roughness of the repair interface is controlled at Ra≤0.1μm, and the bending life reaches 120,000 cycles, exceeding the durability of the original substrate. The system integrates a six-degree-of-freedom robotic arm motion platform and an edge computing unit, achieving 5cm scanning path via a helical involute. 2With a detection coverage efficiency of / s, combined with FPGA acceleration processing, the full inspection time for a single tag is reduced to 30 seconds, a 5.2-fold improvement over traditional manual operation. The cloud-based quality database dynamically optimizes production process parameters through a multivariate regression model, reducing antenna impedance fluctuations from ±3Ω to ±0.5Ω and establishing a defect root cause analysis mechanism, significantly improving production yield to 99.5%. This technology drives a 62% reduction in the overall cost per tag, with annual production capacity exceeding 50 million units. It is not only applicable to RFID tag manufacturing but can also be extended to cutting-edge fields such as flexible sensors and electronic skin, providing a complete technology solution for the large-scale reliable manufacturing of IoT devices.

[0154] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A flexible substrate oriented RFID tag defect intelligent detection system, characterized in that, The system comprises: a high-resolution image acquisition module configured with a ring-shaped light source and an optical microscope lens group, which is used to obtain the micron-level topographic features of the flexible substrate surface, and realizes precise imaging of micro-cracks, wrinkles and surface fracture defects of the conductive layer by eliminating curved surface reflection interference; a multi-spectral sensor array integrating near-infrared and terahertz band sensors, which is used to penetrate the surface layer of the flexible substrate and detect internal conductive layer fracture, delamination and internal bubble defects of the substrate, and enhance the detection ability of hidden defects through multi-band data fusion; a six-degree-of-freedom mechanical arm motion platform carrying the high-resolution image acquisition module and the multi-spectral sensor array, which realizes full-coverage detection of the flexible substrate through a three-dimensional dynamic scanning path, and realizes high-precision defect positioning of large-size substrates by combining sub-pixel level image stitching algorithm; a defect recognition algorithm based on deep learning, which realizes precise classification of various defects including micro-cracks, conductive layer fracture and substrate deformation by combining multi-scale convolutional neural network and transfer learning technology; The system also comprises a dual-mode self-repairing module configured to perform repair according to the classification results of the defect recognition algorithm: for conductive layer defects, nano-silver conductive colloid is injected through micro-fluid channels, and 3D structure reconstruction is realized by using controllable temperature field; for substrate damage, a light-responsive shape memory polymer patch is used, which realizes molecular-level bonding repair after being activated by ultraviolet light.

2. The flexible substrate facing RFID tag defect intelligent detection system of claim 1, wherein, The defect recognition algorithm comprises the following steps: constructing a multi-scale residual convolutional neural network model, the input layer receiving multi-modal features of high-resolution images and multi-spectral data, the backbone network comprising 3 parallel convolution branches with convolution kernel sizes of 3x3, 5x5 and 7x7 respectively, and the receptive field of each branch being expanded by hole convolution; the feature fusion layer dynamically weights the output feature maps of each branch using channel attention mechanism, the pre-trained VGG-16 model weights are loaded in the transfer learning stage and the shallow network parameters are frozen; the output layer calculates the probability distribution of various defects by using Softmax function, and the classification accuracy meets: wherein, is an indicator function, is the total number of test samples, and are the true and predicted labels, respectively.

3. The flexible substrate facing RFID tag defect intelligent detection system of claim 1, wherein, The detection efficiency is improved, which comprises the following steps: Full inspection time for a flexible substrate of size 200x200mm ≤ 30s, 5.2 times faster than traditional manual inspection, and supports minimum 0.1mm defect identification.

4. A method for self-repairing defects of RFID tags facing flexible substrates, which is implemented based on the intelligent detection system for defects of RFID tags facing flexible substrates according to any one of claims 1-3, characterized in that, S1: complete three-dimensional scanning through the mechanical arm motion platform, and synchronously acquire surface topography and internal structure data; S2: classify defects using a multi-scale residual convolutional neural network model, output detection results with a confidence level of ≥95%, and perform conductive layer self-repairing or substrate self-repairing according to the defect classification results using a dual-mode self-repairing mechanism; S3: if it is a conductive layer defect, start the micro-fluid injection-sintering process, and control the temperature error to be ≤±2℃; S4: if it is a substrate damage, match the pre-prepared shape memory polymer from the patch library, and perform ultraviolet light activation bonding; S5: upload the repair data to the cloud quality database through the edge computing unit, update the quality archives and feed back the process optimization parameters. The conductive layer self-repairing method in step S2 comprises:

5. The flexible substrate -facing RFID tag defect self-repairing method of claim 4, wherein, Step 3: detect the mechanical strength of the repaired area by a laser Doppler vibration detector, and ensure that the vibration amplitude error is less than ±5%. Step 1: Defect area positioning through microfluidic channel, based on conductive layer fracture length with defect area Calculate the amount of silver nanocolloid to inject, the volume injected satisfying: wherein R is the radius of the micro-nozzle, 1 .5 is a safety redundancy factor; Step 2: under the protection of nitrogen, the defect area is heated to 120 150℃ by pulse heating, triggering the sintering of nano-silver particles to form a continuous conductive path, and the resistance recovery rate of the defect area is satisfies: wherein, Rpost is the resistance after repair, Rinitial is the initial resistance; The substrate self-repairing method in step S2 comprises:

6. The flexible substrate faced RFID tag defect self-repair method of claim 4, wherein, Step 3: detect the roughness Ra of the repaired interface by an atomic force microscope, and ensure that Ra≤0.1μm. Step 1 : A patch of shape memory polymer is cut using a femtosecond laser, with a thickness matched to the mechanical properties of the substrate, satisfying: wherein, E is the modulus of elasticity of the flexible substrate, E is the modulus of elasticity of the shape memory polymer patch, t is the thickness of the substrate; Step 2: Patch surface azobenzene groups activated by UV light source, triggering molecular chain reconfiguration, bonding time Satisfies: wherein A = 1.2 x 10 3 B = 2.8 x 10 3 is a material constant, is the light intensity, is the wavelength; ​ 7. The flexible substrate faced RFID tag defect self-repair method of claim 4, wherein, The dual-mode self-repairing mechanism in the step S2 automatically switches according to the defect depth: When defect depth d ≥ 0.3 substrate self-repair is initiated, otherwise conductive layer self-repair is performed; After the repair is completed, the tag ID is read through near field communication and compared with historical parameters in a cloud quality database to generate a verification report containing a repair position, material consumption and performance indicators.

8. The flexible substrate -facing RFID tag defect self-repairing method of claim 7, wherein, The radio frequency performance after the repair meets: The tag reading distance recovery rate: wherein, is the post-repair read distance, is the initial read distance, and the resonant frequency offset ≤ ±0.5 MHz.

9. The flexible substrate -facing RFID tag defect self-repairing method of claim 4, wherein, The edge computing unit in the step S5: The integrated FPGA accelerator performs Gaussian filtering and histogram equalization preprocessing on the collected data in real time, and the processing delay meets: wherein, is the number of single frame pixels, is the number of single pixel operations, is the clock frequency.

10. The flexible substrate faced RFID tag defect self-repair method of claim 4, wherein, The cloud quality database in the step S5 realizes the following functions: Store the process parameters, detection results and repair records of each batch of RFID tags, optimize the production parameters through multiple regression analysis, and establish the antenna impedance and the mapping model of printing pressure , temperature : wherein is a regression coefficient, a dynamic adjustment parameter such that - 50 Ω | ≤ 0.5 Ω.

Images