Sensor-based composite delamination detection apparatus and method

Abstract

This invention discloses a sensor-based composite material delamination detection device and method, comprising a controller and a scanning drive assembly, with a detection assembly mounted on the scanning drive assembly. The detection assembly includes a cylindrical shell with an open bottom, an ultrasonic transducer fixedly mounted at the top of the inner part of the cylindrical shell, and a piezoelectric ceramic excitation ring sleeved around the ultrasonic transducer. A flexible acoustic capsule is sealed and fixedly connected to the bottom port of the cylindrical shell, the flexible acoustic capsule having a bowl-shaped elastic film structure with a downward convex center. The sealed cavity in the cylindrical shell is completely filled with a shear-thickening fluid. The controller is electrically connected to the piezoelectric ceramic excitation ring. This invention relates to the field of material detection technology. This invention solves the problems of difficulty in simultaneously achieving flexible adaptive bonding and rigid sound field focusing in existing ultrasonic testing of curved surfaces of composite materials, as well as the problems of complex structure and poor detection consistency caused by the lag in traditional electromechanical servo systems.

Description

Technical Field

[0001] This invention relates to the field of materials testing technology, and in particular to a sensor-based composite material delamination detection device and method. Background Technology

[0002] Composite core components are widely used in high-end equipment due to their advantages such as lightweight, high strength, and high designability. However, their multi-layered structure is prone to closed-loop delamination defects, posing a significant challenge to traditional ultrasonic testing under complex curved surface conditions. Rigid probes struggle to adapt to surface undulations, easily generating lift-off fluctuations and beam deflection, leading to acoustic energy attenuation and focusing failure. While flexible coupling media can improve adhesion, they cause focal drift and wavefront distortion, making it difficult to guarantee test consistency. Existing active electromechanical servo solutions are complex in structure and have slow response times, making it difficult to meet the cycle time and accuracy requirements of efficient online full inspection in core component production lines.

[0003] Especially in typical core material intelligent control scenarios such as high-curvature skins and aerospace composite curved surface components, the detection device needs to simultaneously possess the dual capabilities of "flexible conformal to curved surfaces" and "rigid focusing of sound fields." However, existing technical solutions have long been limited by the physical contradiction that these two capabilities cannot be simultaneously achieved. How to achieve temporal decoupling of flexible bonding in the scanning state and transient focusing in the detection state without the need for complex electromechanical systems has become a key technical bottleneck restricting the intelligent quality control of core material structural components. Summary of the Invention

[0004] To address the challenges of balancing flexible adaptive bonding and rigid acoustic field focusing in existing ultrasonic testing of curved composite surfaces, as well as the problems of complex structures and poor detection consistency caused by lag in traditional electromechanical servo systems, the present invention aims to provide a sensor-based composite material delamination detection device and method.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a sensor-based composite material delamination detection device, comprising a controller and a scanning drive assembly, wherein a detection assembly is tractively mounted on the scanning drive assembly; The detection assembly includes a cylindrical shell with an open bottom. An ultrasonic transducer and a piezoelectric ceramic excitation ring sleeved around the ultrasonic transducer are fixedly installed at the top of the cylindrical shell. The sound beam emitting surface of the ultrasonic transducer faces downward. A flexible sound-transmitting capsule is sealed and fixedly connected to the bottom port of the cylindrical shell. The flexible sound-transmitting capsule has a bowl-shaped elastic film structure with a downward convex center. The sealed cavity in the cylindrical shell is completely filled with a shear-thickening fluid. The controller is electrically connected to the piezoelectric ceramic excitation ring and is used to control the piezoelectric ceramic excitation ring to apply high-frequency shear stress to the shear thickening fluid to trigger a liquid-solid phase transition.

[0006] Preferably, the scanning drive assembly includes a fixedly installed XYZ three-axis combined slide module. An L-shaped plate is fixedly connected to the transmission end of the XYZ three-axis combined slide module. A square rod is slidably connected to the bottom of the L-shaped plate. A baffle that abuts against the top surface of the L-shaped plate is fixedly connected to the top surface of the square rod. The bottom of the square rod is fixedly connected to the top surface of the cylindrical housing. A spring is sleeved on the square rod. The two ends of the spring abut against the top surface of the cylindrical housing 1 and the bottom surface of the L-shaped plate, respectively.

[0007] Preferably, the top surface of the piezoelectric ceramic excitation ring is fixed to the inner top wall of the cylindrical shell, its outer circumferential surface is fixed with the inner wall of the cylindrical shell by interference fit, and its inner circumferential surface is provided with an annular gap of 0.5mm to 2mm between it and the outer wall of the ultrasonic transducer. The lower end face of the piezoelectric ceramic excitation ring is immersed in the shear thickening fluid; the polarization direction of the piezoelectric ceramic excitation ring is perpendicular to the thickness direction, and the resonant frequency is 30kHz to 60kHz.

[0008] Preferably, a circular hole is formed in the middle of the bottom of the flexible acoustic capsule, and an acoustic window is fixedly and sealed to the inner wall of the circular hole. The bottom of the acoustic window is flush with the bottom end of the circular hole. The flexible acoustic capsule is made of high-ductility polyurethane or silicone rubber with a Shore A hardness of 30-50. The acoustic window is made of acoustic impedance matching polyurethane elastomer with a Shore A hardness of 60-80. The acoustic window has a closed-cell gradient porous structure inside, with a closed-cell rate ≥95% and an average pore diameter of 1μm to 30μm. The acoustic window has a dense sealing skin on the side facing the shear-thickening fluid. The thickness of the dense sealing skin is 0.5 μm to 2 μm and there are no through pores. The porosity of the closed-pore gradient porous structure decreases gradually from the dense sealing skin to the bottom surface of the acoustic window, which is used to achieve a continuous transition matching of acoustic impedance and ensure zero leakage of the shear thickening fluid.

[0009] Preferably, an annular plate is fixedly sleeved on the outer wall of the cylindrical shell, and a plurality of infrared thermal imaging modules arranged in an annular array are fixedly installed at the bottom of the annular plate, with the detection end of the infrared thermal imaging module tilted downward toward the acoustic window area.

[0010] Preferably, the shear-thickening fluid is a suspension of nano-silica particles dispersed in polyethylene glycol; the nano-silica particles have a particle size of 100 nm to 300 nm, a mass fraction of 45% to 55%, and a critical shear strain rate threshold for shear-thickening phase transition of 1.5 × 10⁻⁶. 2 s -1 Up to 3.0×10 2 s-1 The shear-thickening fluid has a viscosity of 50 mPa·s to 100 mPa·s in the liquid state and a shear modulus of 50 kPa to 150 kPa in the solid state.

[0011] Preferably, the ultrasonic transducer is a focusing transducer with a focal length of 5mm to 20mm and a center frequency of 1MHz to 5MHz; an acoustic matching layer is provided at the interface between the acoustic beam emitting surface of the ultrasonic transducer and the shear-thickening fluid, and the acoustic impedance value of the acoustic matching layer satisfies the formula: ;in Let be the acoustic impedance of the ultrasonic transducer. The acoustic impedance of the shear-thickening fluid is given.

[0012] Preferably, a backing absorber block is fixedly installed on the top inner side of the rigid shell, and the bottom of the backing absorber block is fixedly connected to the back of the ultrasonic transducer.

[0013] Preferably, an annular heating element is fixedly sleeved on the inner wall of the cylindrical shell, and a plurality of temperature sensors are embedded in an annular array on the inner wall of the cylindrical shell.

[0014] A method of using a sensor-based composite material delamination detection device includes the following steps: S1, the composite material is laid flat on the detection platform, the scanning drive component drives the detection component to move on the surface of the composite material, the piezoelectric ceramic excitation ring is in the closed state, the shear thickening fluid is in a low viscosity liquid state, and the flexible sound-permeable bladder undergoes macroscopic deformation under pressure to passively fit the curved surface of the composite material. S2, within 0.8ms to 1.5ms before the ultrasonic transducer prepares to emit a detection pulse, the controller controls the piezoelectric ceramic excitation ring to open, applying a high-frequency shear stress of 40kHz±5kHz to the shear thickening fluid, causing its shear strain rate to exceed the critical threshold, and the shear thickening fluid to undergo a liquid-solid phase transition, causing the flexible sound-permeable capsule to transiently solidify into a rigid acoustic lens. S3, during the period when the shear-thickening fluid remains in a solidified state, the ultrasonic transducer emits high-frequency focused ultrasonic pulses and receives reflected echoes to complete the sound field locking detection at the current detection point; S4, Sound Field Correction and Defect Feature Extraction: S4a, Establish an acoustic propagation model for shear-thickened fluids to correct acoustic distortion caused by phase transition: ; ;in, For sound pressure, For fluid in time The speed of sound, Shear strain rate This is the critical shear strain rate threshold. and These represent the speeds of sound in liquid and solid fluids, respectively. S4b uses wavelet transform to extract the time-frequency features of the ultrasonic echo: ;in, These are wavelet coefficients. As a scale factor, For translation parameters, c) Extracting defect features from infrared thermograms using the heat conduction equation model: (The provided text appears to be incomplete and contains several errors. A more accurate translation would require the full context.) ;in, For temperature field, Where is the thermal diffusivity, For heat source intensity, For density, Specific heat capacity; S5, Multimodal Feature Fusion and Defect Recognition: S5a, constructing multimodal feature vectors: ;in, This is the ultrasonic time-frequency feature vector. This is the infrared thermal diffusion characteristic vector; S5b, the feature vector Input to the physical constraint neural network model: ; ;in, For network output, It is a convolutional neural network. For data loss, For physical bundle loss, This is the balance coefficient; S5c, the physical constraint bundle loss function is defined as: This loss function forces the network output to conform to the physical laws of heat conduction. S6, Ultrasonic-Infrared Thermometry Co-verification of Closed-Type Delamination Defects: S6a, when the amplitude of the reflected echo signal received in step S3 is lower than the preset threshold, but the defect confidence level output by the network in step S5 is greater than 0.7, the area is marked as a suspected closed-type layered defect area. S6b, in the suspected closed-type delamination defect area, the piezoelectric ceramic excitation ring is kept open, so that the shear thickening fluid is kept in a solidified state to form a rigid coupling; S6c, the ultrasonic transducer switches to a high-power continuous low-frequency ultrasonic excitation mode, injecting high-frequency mechanical vibration into the interior of the composite material, causing micro-friction and heat generation at the closed-type layered interface. S6d, the infrared thermal imaging module acquires thermal radiation images of the composite material surface in real time, and the phase-locked thermal wave processing module in the controller extracts the temperature gradient characteristics of the friction hot spot according to the physical constraint model in step S5. S6e, the depth d and area A of the closed-type layered defect are quantitatively evaluated using the following physical model: ; ;in, For the maximum temperature rise, The average temperature. The ultrasonic excitation frequency, and For calibration coefficients, and This refers to the thermal response time range.

[0015] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: 1. This invention achieves controllable liquid-solid phase change timing switching through the synergistic effect of a piezoelectric ceramic excitation ring and a shear-thickening fluid. During the scanning phase, the low viscosity of the fluid is used in conjunction with a flexible acoustic capsule to passively fit the complex curved surface, eliminating lift-off fluctuations and sound beam deflection. At the moment of detection, the fluid is triggered to transiently solidify and form a rigid acoustic lens to lock the vertical sound field. This completely decouples the physical contradiction between adaptive surface fitting and high-precision acoustic focusing, significantly improving detection consistency and signal-to-noise ratio.

[0016] 2. This invention uses a composite design of closed-pore gradient porous structure and dense sealing skin to make the porosity inside the acoustic window decrease in a gradient to achieve a continuous and smooth transition of acoustic impedance. At the same time, the dense skin without through pores blocks the fluid penetration path, ensuring efficient transmission of ultrasonic waves while achieving zero leakage of the coupling medium, effectively extending the fatigue life and sealing reliability of the probe assembly.

[0017] 3. This invention constructs a multimodal feature vector that integrates ultrasonic time-frequency features and infrared thermal diffusion features, and introduces a physical constraint loss function based on the residual of the thermal conduction equation to train the neural network. This forces the model output to follow the real physical thermal diffusion law, effectively overcoming the overfitting defect of pure data-driven algorithms when samples are scarce, and improving the confidence of defect classification and the physical interpretability of quantitative inversion of depth and area. Attached Figure Description

[0018] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments: Figure 1 This is a schematic diagram of the overall side view of the present invention; Figure 2 This is a cross-sectional structural diagram of the present invention.

[0019] In the diagram: 100, Detection component; 1, Cylindrical housing; 2, Ultrasonic transducer; 3, Piezoelectric ceramic excitation ring; 4, Flexible acoustic capsule; 41, Acoustic window; 5, Infrared thermal imaging module; 51, Ring plate; 200, Scanning drive component; 201, XYZ three-axis combined slide module; 202, L-shaped plate; 203, Square rod; 204, Baffle; 205, Spring. Detailed Implementation

[0020] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification.

[0021] Please see Figures 1 to 2 It should be understood that the structures, proportions, sizes, etc., illustrated in the accompanying drawings are merely for illustrative purposes to aid those skilled in the art and to facilitate understanding and reading. They are not intended to limit the scope of the invention and therefore have no substantial technical significance. Any modifications to the structure, changes in proportions, or adjustments to size, without affecting the effectiveness and purpose of the invention, should still fall within the scope of the technical content disclosed in this invention. Furthermore, the terms such as "upper," "lower," "left," "right," "middle," and "one" used in this specification are merely for clarity and not intended to limit the scope of the invention. Changes or adjustments to their relative relationships, without substantially altering the technical content, should also be considered within the scope of the invention's implementation.

[0022] Example 1: Overall mechanical structure and assembly relationship of the detection device. This example provides a sensor-based composite material delamination detection device, which mainly includes a controller, a scanning drive component 200 and a detection component 100.

[0023] The scanning drive assembly 200 adopts an XYZ three-axis combined slide module 201, with its Z-axis drive end fixedly connected to an L-shaped plate 202 via bolts. A square through-hole is formed at the bottom of the L-shaped plate 202, through which a square rod 203 slides. A limiting baffle 204 is fixedly welded to the top of the square rod 203, with its upper surface abutting against the top surface of the L-shaped plate 202 to prevent the square rod from dislodging. The bottom end of the square rod 203 is rigidly connected to the top surface of the cylindrical housing 1 via a flange. A compression spring 205 is fitted in the middle section of the square rod 203, with its upper end abutting against the bottom surface of the L-shaped plate 202 and its lower end abutting against the top surface of the cylindrical housing 1. During assembly, the spring 205 is pre-compressed to 30%~50% of its stroke, providing a constant downward pressure of 5N~15N, enabling the detection assembly 100 to have flexible floating capability in the vertical direction during scanning.

[0024] The core of the detection component 100 is a cylindrical housing 1, preferably made of aerospace-grade aluminum alloy 6061-T6 or carbon fiber composite material. An ultrasonic transducer 2 is fixedly mounted inside the top of the housing 1 using epoxy structural adhesive, with the sound beam emitting surface of the transducer 2 strictly facing downwards. A piezoelectric ceramic excitation ring 3 is coaxially sleeved around the outer periphery of the transducer 2, its top surface fixed to the inner top wall of the housing 1 by a pressure ring. The outer circumferential surface and the inner wall of the housing are thermo-pressed with an interference fit, while the inner circumferential surface maintains a 1.0mm annular gap with the outer wall of the transducer 2. This gap is filled with a polyurethane damping ring to isolate structural sound transmission. The lower end of the excitation ring 3 is immersed in a shear-thickening fluid 8, with the immersion depth controlled at 5mm ± 0.5mm. The excitation ring 3 uses PZT-5H piezoelectric ceramic, polarized along the thickness direction, with a measured resonant frequency of 40.2kHz.

[0025] The outer wall of the cylindrical housing 1 is fixedly fitted with an annular plate 51 by threads. Six infrared thermal imaging modules 5 are evenly distributed at the bottom of the annular plate 51 in a ring array. The optical axis of each module 5 is tilted inward at a 25° angle to the central axis of the housing 1. The working distance of the lens is set to 80 mm, and the field of view is 28°×21°. This ensures that after the line of sight passes through the infrared transmission window on the side wall of the housing, it focuses on the annular area 1.5 mm to 2.5 mm outside the contact area between the bottom surface of the acoustic window 41 and the workpiece.

[0026] A backing absorber block 11, made of tungsten powder / epoxy resin composite material, is bonded to the inner top of the cylindrical housing 1. It is 8mm thick and has a diameter larger than the outer diameter of the transducer 2. This absorber block is used to absorb back-facing sound waves and dampen residual vibrations of the piezoelectric crystal. A flexible annular heating element 6 is circumferentially attached to the inner wall of the housing 1, and three miniature PT100 temperature sensors 7 are evenly distributed along the circumference. The sensor probes extend 3mm into the fluid 8, and a PID closed-loop algorithm maintains the fluid operating temperature at 25℃±1℃.

[0027] Example 2: Materials and preparation of flexible acoustic capsules, acoustic windows, and shear-thickening fluids: The flexible acoustic capsule 4 is made of liquid silicone rubber (LSR) with a Shore hardness of A40, precision molded into a dome shape with a Φ18mm circular hole at the center of the bottom. The capsule 4 has an edge width of 4mm and an O-ring sealing groove machined on the inner side. It is double-sealed at the bottom port of the cylindrical shell 1 by a fluororubber sealing ring and anaerobic adhesive. It showed no leakage after a 0.2MPa pressure holding test for 30 minutes.

[0028] The acoustic window 41 is nested and fixed within the circular hole, with its bottom surface flush with the bottom surface of the capsule 4. The window is made of polyurethane elastomer with a Shore A hardness of 70, and is fabricated using a micro-foaming gradient molding + surface plasma densification process. The polyurethane prepolymer is mixed with a physical foaming agent and injected into a mold with a temperature gradient, with the top temperature at 80°C and the bottom temperature at 60°C. The mold is then pressed and cured to form an elastomer preform with an internal closed-cell rate of 96%, and the porosity gradually changes from 18% on the top surface to 4% on the bottom surface, with an average pore size of 15μm.

[0029] The top surface of the contact fluid 8 is treated with low-temperature oxygen plasma to cross-link the surface molecular chains and form a dense sealing skin layer with a thickness of 1.2 μm without through pores.

[0030] Tested with a helium mass spectrometer, the leakage rate was ≤8×10⁻⁶ under a static pressure of 0.15 MPa. -9 Pa·m³ / s, meeting the zero leakage requirement. This gradient structure allows the acoustic impedance to smoothly transition from 1.15 MRayl at the top to 1.48 MRayl at the bottom, with an impedance mismatch of <8% with the shear-thickening fluid 8 and the carbon fiber composite material.

[0031] The shear-thickening fluid 8 was prepared by dispersing nano-silica in polyethylene glycol PEG-400 using a fumed silica method. The silica mass fraction was 50%, and the fluid was packaged after undergoing high-speed shear dispersion and vacuum degassing. The liquid viscosity at room temperature was 75 mPa·s, and the critical strain rate threshold for the shear-thickening phase transition was 2.0 × 10⁻⁶. 2 s -1 The measured solid-state shear modulus after phase transition is 85 kPa. The ultrasonic transducer 2's emitting surface is coated with an acoustic matching layer, and its acoustic impedance is calculated according to the formula... ;in The acoustic impedance of the ultrasonic transducer 12 is given by [reference to specific parameters]. Acoustic impedance of the shear-thickening fluid (14) Example 3: Implementation process of hierarchical detection method and physical constraint neural network model: This embodiment details the detection method based on the above-described device, with specific steps and algorithm implementation as follows: S1. Flexible Adaptive Scanning: The scanning drive component 200 drives the detection component 100 to move along the surface of the composite material to be tested. The piezoelectric ceramic excitation ring 3 is de-energized and shut off, and the shear thickening fluid 8 is in a low-viscosity liquid state. Under the constant pressure of the spring 205, the flexible acoustic capsule 4 undergoes macroscopic elastic deformation, passively adhering to the workpiece surface. The acoustic window 41 maintains normal contact with the surface, and the lift-off fluctuation is controlled within ±0.15mm.

[0032] S2. Transient acoustic field locking: 1.2 ms before the ultrasonic transducer 2 emits the detection pulse, the controller applies a 40 kHz / 60 Vpp sinusoidal alternating voltage to the excitation ring 3. The excitation ring generates tangential high-frequency micro-vibrations with an amplitude of 3.5 μm, causing the local shear strain rate of the fluid to instantaneously jump to 2.8 × 10⁻⁶. 2 s -1 Exceeding the critical threshold triggers a liquid-solid phase transition. The flexible, acoustically transparent capsule 4 transforms from flexible to rigid within microseconds, and the sound velocity... leap to This forms a transient focusing acoustic lens.

[0033] S3. Ultrasonic Pulse Detection: After the phase transition is complete and the sound field stabilizes, transducer 2 emits a 2.25MHz / 3-cycle high-voltage pulse and receives the reflected echo. The acquisition card records the A-scan signal at a sampling rate of 100MS / s, completing the sound field locking detection with a step size of 0.5mm at the current detection point.

[0034] S4. Sound Field Correction and Defect Feature Extraction: S4a Acoustic Propagation Model Correction: The controller's built-in DSP chip solves the acoustic wave equation in real time. The finite difference method (FDTD) is used to discretize the wave equation, and the core equation is as follows: The algorithm is set with a grid step size Δx = 0.1 mm and a time step size Δt = 50 ns. The system can dynamically switch the sound propagation coefficient according to the real-time driving state of the excitation ring. The value of is precisely used to compensate for the sound path drift caused by the phase change of the medium, and the ultrasonic echo positioning error is ultimately controlled within the range of <0.08mm, ensuring the accuracy of sound field propagation.

[0035] S4b wavelet time-frequency feature extraction: For the acquired ultrasonic echo signals Continuous wavelet transform (CWT) processing was performed, using complex Morlet wavelets as the basis functions, with a center frequency of 1 Hz and a bandwidth parameter of 1.5. The scaling factor was set to... After multi-scale signal decomposition, 12-dimensional ultrasonic features, such as the energy ratio of high-frequency components, the main frequency offset, and the waveform entropy, are accurately extracted to construct an ultrasonic feature vector. .

[0036] S4c Infrared Thermal Conductivity Feature Extraction: The infrared module (5) continuously acquires thermal radiation images of the area under test at a frame rate of 30fps, and constructs a spatiotemporal temperature matrix. The heat conduction equation is solved discretically using the alternating direction implicit method (ADI). The core heat conduction equation is as follows: The thermal diffusivity is obtained through equation inversion calculation. With equivalent heat source intensity Based on the spatiotemporal distribution pattern of the hotspot, eight-dimensional infrared thermal features such as peak temperature, temperature gradient amplitude, and thermal diffusion radius are further extracted to construct an infrared feature vector. .

[0037] S5. Multimodal feature fusion and defect identification: S5a Feature Vector Construction: Ultrasonic features With infrared thermal characteristics After performing dimensional concatenation and normalization preprocessing to eliminate the influence of dimensions, a 20-dimensional standardized multimodal feature vector is finally constructed, as shown in the following expression: ; S5b Physically Constrained Neural Network PINN Inference: The constructed multimodal feature vector The pre-trained convolutional neural network model is input, employing a 3-layer convolutional structure with 3×3 kernels and channel numbers set to 32, 64, and 128 respectively. All layers use the ReLU activation function and are then connected to a fully connected layer after global average pooling. The network ultimately outputs a defect probability map and defect depth regression values, enabling preliminary defect localization and quantitative identification.

[0038] The model training uses a combined loss function, and the overall loss formula is: ;in, This is a hybrid data loss consisting of cross-entropy loss and MSE loss, used to fit sample label information; physical loss weighting coefficients. The optimal value is determined using a grid search algorithm.

[0039] S5c Physical Constraint Loss Calculation: During the model training phase, a physical constraint loss is introduced, and the core calculation formula is as follows: By employing automatic differentiation technology, the partial derivatives of the network's predicted output with respect to time and space are accurately calculated, forcing the model's predicted thermal diffusion field to strictly adhere to Fourier's law of heat conduction. This physical constraint mechanism effectively improves model performance in small-sample scenarios. Under the condition of fewer than 500 labeled data sets, the model's generalization error is reduced by 42%, while ensuring that various physical quantities in the network output satisfy the conservation law, thus enhancing the physical rationality of the prediction results.

[0040] S6. Ultrasonic-infrared thermographic co-examination of closed-type delamination defects: S6a~S6b Suspected Defect Area Identification: System judgment logic: When the ultrasonic echo amplitude detected by the S3 link is < -40dB, it is judged as a suspected closed delamination defect feature. At the same time, when the defect recognition confidence of the S5 neural network is > 0.7, the corresponding area is marked as a suspected defect area. After the judgment is completed, the controller keeps the excitation ring 3 open to maintain the rigid coupling state of fluid solidification, providing a stable detection condition for subsequent collaborative verification.

[0041] S6c~S6d Thermal Wave Excitation and Image Acquisition Noise Reduction: The transducer (2) is switched to a 20kHz, 200Vpp continuous excitation mode and continuously excited for 1.5s. Vibration excitation is used to generate a micro-frictional heating effect at the closed layered interface of the component, while the infrared module (5) synchronously and continuously acquires thermal image sequences. The infrared image data is optimized by the phase-locked loop processing module to accurately extract the core features of the temperature gradient, effectively eliminate external interference such as environmental reflection and temperature noise, and ensure the validity of the thermal feature data.

[0042] S6e Defect Quantitative Assessment and Alarm Output: The system uses dual calibration coefficients for quantitative defect calculation, and the calibration parameters are: , Substitute the values ​​into the corresponding algorithm formula to solve for the defect depth. With the projected area of ​​the defect .

[0043] Defect Judgment and Output Rules: When the detection result meets the following conditions... and When the defect is detected, the system automatically triggers a defect alarm and accurately marks the spatial coordinates of the defect, thus completing the accurate detection and quantitative calibration of closed-type layered defects.

[0044] Example 4: Key Parameter Calibration and Algorithm Training Implementation: To ensure that those skilled in the art can fully reproduce the algorithm and quantitative detection model proposed in this invention, and to guarantee the feasibility and repeatability of the technical solution, supplementary explanations are provided regarding the calibration of core physical parameters, the construction of the PINN model dataset, the model training strategy, and the quantitative coefficient calibration process. The specific implementation process is as follows.

[0045] 4.1 Calibration of thermal diffusivity α: The in-plane thermal diffusivity of standard carbon fiber / epoxy resin laminate samples was tested using the laser flare method (LFA). Multiple sets of thermal response data were collected using high-precision thermophysical testing equipment. After outliers were removed, statistical fitting was performed, and the measured baseline value of the material's thermal diffusivity was obtained. This parameter, as a fixed prior parameter, is input into the S4c heat conduction equation solving module and the S5c physical constraint loss calculation module, providing accurate intrinsic material parameters to support thermal field inversion and physical constraint modeling, ensuring the realism and accuracy of the thermal diffusion field calculation.

[0046] 4.2 Construction of PINN Model Training Dataset and Training Strategy: To address the issues of scarce and unevenly distributed defect samples in practical engineering, a full-process multiphysics coupled simulation model was built using the COMSOL Multiphysics simulation platform, encompassing "ultrasonic excitation - interfacial frictional heat generation - heat diffusion and transfer." By traversing different defect conditions and parameters, simulation sample data was generated in batches, covering various scenarios including delamination depths of 0.1–3 mm, defect projected areas of 2–50 mm², and different composite material layup angles. A total of 2000 pairs of one-to-one simulated infrared thermograms and ultrasonic echo signals were generated, covering the entire effective range of the detection conditions described in this invention.

[0047] The constructed sample dataset was randomly divided in an 8:1:1 ratio to correspond to the model training set, validation set, and test set, ensuring the rationality of the dataset division and the objectivity of the model evaluation. The model was trained using the PyTorch deep learning framework, with a total of 150 epochs. The Adam adaptive optimizer was used to update the parameters iteratively, and the initial learning rate was set to 1e-3.

[0048] To balance data fitting accuracy and the rationality of physical constraints, and to improve the stability of small-sample training, physical constraint weights are... An adaptive dynamic adjustment strategy is adopted: the weight is set to 0.1 in the early stage of training to prioritize fitting the features of the sample data and ensure the basic convergence of the model; in the later stage of training, the weight is increased to 0.3 to strengthen the constraints of the physical equations, correct the model fitting bias, and finally achieve the optimal combination of data-driven and physical prior.

[0049] 4.3 Calibration of quantitative model coefficients k1 and k2: To determine the core correction coefficient for the quantitative defect calculation formula, a standard composite material test block with an embedded polytetrafluoroethylene (PTFE) film was fabricated. The PTFE film simulated real closed-type delamination defects within the component, closely matching the defect characteristics of the actual test object. A standardized and fixed ultrasonic excitation frequency was used. Consistent with the S6 inspection conditions, the maximum temperature rise of multiple sets of artificial defects of different sizes was collected. Average temperature And the continuous thermal response integral curve.

[0050] Based on the measured true values ​​of defect depth and area and the theoretical calculation values ​​of the algorithm, the least squares method is used for global fitting optimization, the quantitative formula correction coefficients are iteratively corrected, and the optimal coefficients are finally calibrated. , The above coefficients are applicable to carbon fiber / epoxy resin composite systems. For other material systems, the coefficient parameters only need to be refitted through the same calibration process. The main structure of the quantitative calculation formula does not need to be modified, and the model has good versatility and scalability.

[0051] Example 5, Symbol and Parameter Explanation To facilitate those skilled in the art to clearly, accurately, and completely understand the technical solutions, mathematical models, and algorithm logic described in claims 7 and 10 of this invention, to clarify the physical meaning and value specifications of each formula and operator, and to ensure that the technical solutions of this invention meet the requirements for full patent disclosure, all letter symbols, mathematical operators, parameter units, and technical definitions involved in this invention are hereby uniformly explained as follows.

[0052] I. Acoustic Propagation and STF Phase Transition Model

[0053] II. Acoustic Impedance Matching Model

[0054] III. Signal Processing and Wavelet Transform

[0055] IV. Thermal conduction and infrared feature extraction

[0056] V. Multimodal Fusion and Physically Constrained Neural Networks

[0057] VI. Quantitative Assessment Model for Closed-Type Layered Defects

[0058] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. A sensor-based composite material delamination detection device, comprising a controller and a scanning drive assembly (200), characterized in that: The scanning drive assembly (200) is equipped with a detection assembly (100). The detection assembly (100) includes a cylindrical shell (1) with an open bottom. An ultrasonic transducer (2) and a piezoelectric ceramic excitation ring (3) are fixedly installed at the top of the cylindrical shell (1) and sleeved on the outer periphery of the ultrasonic transducer (2). The sound beam emitting surface of the ultrasonic transducer (2) is arranged facing downward. A flexible sound-permeable capsule (4) is sealed and fixedly connected to the bottom port of the cylindrical shell (1). The flexible sound-permeable capsule (4) has a bowl-shaped elastic film structure with a downward protrusion in the middle. The sealed cavity in the cylindrical shell (1) is completely filled with a shear-thickening fluid. The controller is electrically connected to the piezoelectric ceramic excitation ring (3) and is used to control the piezoelectric ceramic excitation ring (3) to apply high-frequency shear stress to the shear thickening fluid to trigger liquid-solid phase change.

2. The sensor-based composite material delamination detection device according to claim 1, characterized in that: The scanning drive assembly (200) includes a fixedly installed XYZ three-axis combined slide module (201). The transmission end of the XYZ three-axis combined slide module (201) is fixedly connected to an L-shaped plate (202). A square rod (203) is slidably connected to the bottom of the L-shaped plate (202). A baffle (204) that blocks the top surface of the square rod (203) is fixedly connected to the top surface of the L-shaped plate (202). The bottom of the square rod (203) is fixedly connected to the top surface of the cylindrical shell (1). A spring (205) is sleeved on the square rod (203). The two ends of the spring (205) are respectively pressed against the top surface of the cylindrical shell (1) and the bottom surface of the L-shaped plate (202).

3. The sensor-based composite material delamination detection device according to claim 1, characterized in that: The top surface of the piezoelectric ceramic excitation ring (3) is fixed to the inner top wall of the cylindrical shell (1), and its outer circumferential surface is fixed to the inner wall of the cylindrical shell (1) with an interference fit. A ring gap of 0.5 mm to 2 mm is provided between its inner circumferential surface and the outer wall of the ultrasonic transducer (2). The lower end face of the piezoelectric ceramic excitation ring (3) is immersed in the shear thickening fluid; the polarization direction of the piezoelectric ceramic excitation ring (3) is perpendicular to the thickness direction, and the resonant frequency is 30kHz to 60kHz.

4. The sensor-based composite material delamination detection device according to claim 1, characterized in that: The flexible acoustic capsule (4) has a circular hole in the middle of its bottom, and an acoustic window (41) is fixedly and sealed to the inner wall of the circular hole. The bottom of the acoustic window (41) is flush with the bottom port of the circular hole. The flexible acoustic capsule (4) is made of high-ductility polyurethane or silicone rubber with a Shore A hardness of 30-50. The acoustic window (41) is made of acoustic impedance matching polyurethane elastomer with a Shore A hardness of 60-80. The acoustic window (41) has a closed-pore gradient porous structure inside, with a closed-pore rate ≥95% and an average pore diameter of 1μm to 30μm. The acoustic window (41) has a dense sealing skin on the side facing the shear-thickening fluid. The thickness of the dense sealing skin is 0.5 μm to 2 μm and there are no through pores. The porosity of the closed-pore gradient porous structure decreases gradually from the dense sealing skin to the bottom surface of the acoustic window (41) to achieve a continuous transition matching of acoustic impedance, while ensuring zero leakage of the shear thickening fluid.

5. The sensor-based composite material delamination detection device according to claim 1, characterized in that: The outer wall of the cylindrical shell (1) is fixedly fitted with an annular plate (51), and a number of infrared thermal imaging modules (5) arranged in an annular array are fixedly installed at the bottom of the annular plate (51). The detection end of the infrared thermal imaging module (5) is tilted downward toward the acoustic window (41) area.

6. The sensor-based composite material delamination detection device according to claim 1, characterized in that: The shear-thickening fluid is a suspension of nano-silica particles dispersed in polyethylene glycol; the nano-silica particles have a particle size of 100 nm to 300 nm, a mass fraction of 45% to 55%, and a critical shear strain rate threshold for shear-thickening phase transition of 1.5 × 10⁻⁶. 2 s -1 Up to 3.0×10 2 s -1 The shear-thickening fluid has a viscosity of 50 mPa·s to 100 mPa·s in the liquid state and a shear modulus of 50 kPa to 150 kPa in the solid state.

7. The sensor-based composite material delamination detection device according to claim 1, characterized in that: The ultrasonic transducer (2) is a focusing transducer with a focal length of 5mm to 20mm and a center frequency of 1MHz to 5MHz; an acoustic matching layer is provided at the interface between the acoustic beam emitting surface of the ultrasonic transducer (2) and the shear-thickening fluid, and the acoustic impedance value of the acoustic matching layer satisfies the formula: ;in The acoustic impedance of the ultrasonic transducer (12) is given by [the value of the transducer]. The acoustic impedance of the shear-thickening fluid (14) is given.

8. A sensor-based composite material delamination detection device according to claim 1, characterized in that: A backing absorber block (11) is fixedly installed on the inner top of the rigid housing (1), and the bottom of the backing absorber block (11) is fixedly connected to the back of the ultrasonic transducer (12).

9. A sensor-based composite material delamination detection device according to claim 1, characterized in that: The inner wall of the cylindrical shell (1) is fixedly fitted with an annular heating element (6), and the inner wall of the cylindrical shell (1) is embedded with a number of temperature sensors (7) in an annular array.

10. The method of using the sensor-based composite material delamination detection device according to claim 1, characterized in that: S1, the composite material is laid flat on the detection platform, the scanning drive component (200) drives the detection component (100) to move on the surface of the composite material, the piezoelectric ceramic excitation ring (3) is in the closed state, the shear thickening fluid is in a low viscosity liquid state, and the flexible sound-permeable bladder (4) is subjected to pressure and undergoes macroscopic deformation to passively fit the curved surface of the composite material; S2, within 0.8ms to 1.5ms before the ultrasonic transducer (2) prepares to emit a detection pulse, the controller (100) controls the piezoelectric ceramic excitation ring (3) to open, and applies a high-frequency shear stress of 40kHz±5kHz to the shear thickening fluid, so that its shear strain rate exceeds the critical threshold, and the shear thickening fluid undergoes a liquid-solid phase change, so that the flexible sound-permeable capsule (4) is transiently solidified into a rigid acoustic lens; S3, during the period when the shear-thickening fluid remains in a solidified state, the ultrasonic transducer (2) emits high-frequency focused ultrasonic pulses and receives reflected echoes to complete the sound field locking detection at the current detection point; S4, Sound Field Correction and Defect Feature Extraction: S4a, Establish an acoustic propagation model for shear-thickened fluids to correct acoustic distortion caused by phase transition: ; ;in, For sound pressure, For fluid in time The speed of sound, Shear strain rate This is the critical shear strain rate threshold. and These represent the speeds of sound in liquid and solid fluids, respectively. S4b uses wavelet transform to extract the time-frequency features of the ultrasonic echo: ;in, These are wavelet coefficients. As a scale factor, For translation parameters, c) Extracting defect features from infrared thermograms using the heat conduction equation model: (The provided text appears to be incomplete and contains several errors. A more accurate translation would require the full context.) ;in, For temperature field, Where is the thermal diffusivity, For heat source intensity, For density, Specific heat capacity; S5, Multimodal Feature Fusion and Defect Recognition: S5a, constructing multimodal feature vectors: ;in, This is the ultrasonic time-frequency feature vector. This is the infrared thermal diffusion characteristic vector; S5b, the feature vector Input to the physical constraint neural network model: ; ;in, For network output, It is a convolutional neural network. For data loss, For physical bundle loss, This is the balance coefficient; S5c, the physical constraint bundle loss function is defined as: This loss function forces the network output to conform to the physical laws of heat conduction. S6, Ultrasonic-Infrared Thermometry Co-verification of Closed-Type Delamination Defects: S6a, when the amplitude of the reflected echo signal received in step S3 is lower than the preset threshold, but the defect confidence level output by the network in step S5 is greater than 0.7, the area is marked as a suspected closed-type layered defect area. S6b, in the suspected closed-type delamination defect area, the piezoelectric ceramic excitation ring (3) is kept open, so that the shear thickening fluid is kept in a solidified state to form a rigid coupling; S6c, the ultrasonic transducer (2) switches to a high-power continuous low-frequency ultrasonic excitation mode, injecting high-frequency mechanical vibration into the interior of the composite material, causing micro-friction and heat generation at the closed-type layered interface. S6d, the infrared thermal imaging module (6) acquires thermal radiation images of the composite material surface in real time, and the phase-locked thermal wave processing module (101) in the controller (100) extracts the temperature gradient characteristics of the friction hot spot according to the physical constraint model in step S5. S6e, the depth d and area A of the closed-type layered defect are quantitatively evaluated using the following physical model: ; ;in, For the maximum temperature rise, The average temperature. The ultrasonic excitation frequency, and For calibration coefficients, and This refers to the thermal response time range.

Images