Intelligent detection method and system for anisotropy of bulletproof material

Abstract

The invention discloses a bulletproof material anisotropy intelligent detection method and system, and relates to the technical field of intelligent detection.The method comprises the steps that S1, a piezoelectric fiber composite material transducer array is arranged on the surface of a bulletproof material to be detected, synchronously emitting two beams of shear waves with the same frequency, orthogonal polarization directions and programmable phase difference to the bulletproof material to be detected, and forming an interference field in the bulletproof material to be detected; s2, performing time-frequency decomposition on the interference signal to extract spatial distribution of an interlayer entropy yield, and calculating and determining interlayer phase field distribution according to a thermodynamic conjugate relationship between the interlayer entropy yield and an interlayer damage degree; s3, based on the interlayer phase field distribution and the fiber orientation gradient of the bulletproof material to be detected, utilizing the damage evolution model to deduce predicted spatial distribution of the interlayer phase field distribution at the next moment; and S4, adjusting the excitation phase difference and the ultrasonic intensity of the shear wave in real time according to the spatial gradient of the predicted spatial distribution, and repeatedly executing the steps S1 to S3 based on the adjusted parameters.

Description

Technical Field

[0001] This application relates to the field of intelligent detection technology, and more specifically, to an intelligent detection method and system for anisotropy of bulletproof materials. Background Technology

[0002] As a core component of critical protective equipment, the performance of ballistic materials directly affects personal safety and protective effectiveness. In practical use, ballistic materials often face complex stress environments and impact loads, and the anisotropic characteristics of their internal structure and interlaminar state significantly influence overall ballistic performance. Therefore, developing effective testing methods to assess their internal structure and state is crucial for ensuring material reliability, optimizing protective design, and extending service life.

[0003] Currently, while testing methods for bulletproof materials can detect obvious macroscopic defects, they still face a fundamental challenge: accurately predicting and assessing the performance evolution of bulletproof materials under actual dynamic loads using test signals. This limitation means that existing testing methods often remain at the level of post-event verification, failing to provide forward-looking guidance for the condition management of bulletproof materials.

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This application provides an intelligent detection method and system for anisotropy of bulletproof materials to solve the above-mentioned technical problems.

[0006] This application provides an intelligent detection method for anisotropy of bulletproof materials, comprising: S1, arranging a piezoelectric fiber composite material transducer array on the surface of the bulletproof material to be tested, synchronously emitting two shear waves with the same frequency, orthogonal polarization direction, and programmable adjustable phase difference into the bulletproof material to be tested, forming an interference field inside the bulletproof material to be tested, and acquiring the interference signal of the interference field; S2, performing time-frequency decomposition on the interference signal to extract the spatial distribution of interlayer entropy yield, and calculating and determining the interlayer phase field distribution based on the thermodynamic conjugate relationship between the interlayer entropy yield and the interlayer damage degree; S3, based on the interlayer phase field distribution and the fiber orientation gradient of the bulletproof material to be tested, using a damage evolution model including a geometrically necessary orientation gradient dissipation term, deducing the predicted spatial distribution of the interlayer phase field distribution at the next moment; S4, adjusting the excitation phase difference and ultrasonic intensity of the shear waves in real time according to the spatial gradient of the predicted spatial distribution, and repeating S1 to S3 based on the adjusted excitation phase difference and ultrasonic intensity.

[0007] This application provides an intelligent detection system for anisotropy of bulletproof materials, comprising: a data acquisition unit, configured to arrange a piezoelectric fiber composite material transducer array on the surface of the bulletproof material to be tested, synchronously emit two shear waves with the same frequency, orthogonal polarization directions, and programmable adjustable phase difference into the bulletproof material to be tested, forming an interference field inside the bulletproof material to be tested, and acquiring the interference signal of the interference field; a calculation unit, configured to perform time-frequency decomposition on the interference signal to extract the spatial distribution of interlayer entropy yield, and calculate and determine the interlayer phase field distribution based on the thermodynamic conjugate relationship between the interlayer entropy yield and the interlayer damage degree; a deduction unit, configured to deduce the predicted spatial distribution of the interlayer phase field distribution at the next time step based on the interlayer phase field distribution and the fiber orientation gradient of the bulletproof material to be tested, using a damage evolution model including a geometrically necessary orientation gradient dissipation term; and an adjustment unit, configured to adjust the excitation phase difference and ultrasonic intensity of the shear waves in real time according to the spatial gradient of the predicted spatial distribution.

[0008] Based on the embodiments provided in this application, two shear waves of a specific form are emitted simultaneously to form a controllable interference field inside the material, thereby actively exciting a mechanical response closely related to anisotropy. This provides a raw data basis for detection, including rich directional information, and effectively improves the excitation efficiency and capture sensitivity of the anisotropic characteristics of bulletproof materials.

[0009] Building upon this foundation, by performing time-frequency decomposition on the interference signal and extracting the interlayer entropy yield, and then calculating the interlayer phase field distribution based on thermodynamic principles, this method transforms the detection signal into physical parameters that directly characterize the energy dissipation and damage state within the material. This facilitates in-depth identification and quantitative assessment of early microscopic damage at critical locations such as interlayer interfaces. Furthermore, by combining the material's inherent fiber orientation gradient and using a damage evolution model that includes geometrically necessary orientation gradient dissipation terms, a predictive spatial distribution of the damage state can be obtained. This enables the detection method to dynamically predict the damage evolution trend of the material under stress, providing a forward-looking basis for assessing its remaining protective performance. By adjusting the excitation parameters in real time based on the predicted distribution and repeating the above process, a closed loop of detection, analysis, prediction, and adjustment that adapts to changes in the material's state is formed, significantly improving the overall detection method's relevance, intelligence, and assessment reliability in complex working conditions. Attached Figure Description

[0010] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a flowchart of an optional intelligent detection method for anisotropy of bulletproof materials according to an embodiment of this application; Figure 2 This is a flowchart of another optional intelligent detection method for anisotropy of bulletproof materials according to an embodiment of this application; Figure 3 This is a flowchart of another optional intelligent detection method for anisotropy of bulletproof materials according to an embodiment of this application; Figure 4 This is a structural diagram of an optional intelligent detection system for anisotropy of bulletproof materials according to an embodiment of this application; Figure 5 This is a schematic diagram of the structure of an optional electronic device according to an embodiment of this application.

[0011] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0013] According to one aspect of the embodiments of this application, such as Figure 1 As shown, this application provides an intelligent detection method for the anisotropy of bulletproof materials, including: S1. A piezoelectric fiber composite material transducer array is arranged on the surface of the bulletproof material to be tested. Two shear waves with the same frequency, orthogonal polarization direction and programmable phase difference are simultaneously emitted into the bulletproof material to be tested, forming an interference field inside the bulletproof material to be tested, and the interference signal of the interference field is collected. When implementing this testing method, a piezoelectric fiber composite transducer array must first be arranged on the surface of the bulletproof material to be tested. Considering that bulletproof helmets or bulletproof plates usually have complex curved surface geometries, and that interlayer damage often occurs within a specific depth range in the thickness direction, this embodiment adopts a flexible attached array arrangement.

[0014] Specifically, the array elements adopt a modular design, with each module containing two orthogonally polarized shear wave emitting units, corresponding to two shear waves with the same frequency but orthogonal polarization directions. The determination of the element spacing needs to comprehensively consider both detection resolution and acoustic attenuation characteristics: for bulletproof laminates made of typical aramid fibers or ultra-high molecular weight polyethylene fibers, when it is necessary to resolve millimeter-level interlaminar defects, the center-to-center spacing of the array elements should be set to one-third to one-half of the shear wave wavelength at the depth of interest to ensure that spatial sampling meets the Nyquist criterion while avoiding excessive coupling between array elements.

[0015] For irregular surfaces like helmets with hyperboloid features, the array employs a segmented adaptive attachment strategy. The array is divided into several flexible subarrays, each of which is attached to the helmet surface via an elastic coupling layer (such as a silicone rubber pad). A local curvature compensation algorithm is used to adjust the spatial orientation of each subarray, ensuring that the transducer radiating surface maintains a reasonable contact angle with the material surface normal, thereby guaranteeing that the shear wave effectively couples into the material interior in the expected mode.

[0016] The phase difference between two orthogonally polarized shear waves is adjusted in real time via a programmable delay line, forming a controllable interference field within the material. The spatial distribution of this interference field is modulated by changes in the local mechanical properties of the material. The interference signal is acquired by receiving units in an array, providing raw data for subsequent interlayer state characterization.

[0017] S2, perform time-frequency decomposition on the interference signal to extract the spatial distribution of interlayer entropy yield, and calculate and determine the interlayer phase field distribution based on the thermodynamic conjugate relationship between interlayer entropy yield and interlayer damage degree; Interlaminar entropy production rate is the core physical quantity in this detection method, used to quantitatively characterize the thermodynamic irreversible dissipation process at the interlaminar interface of bulletproof materials. In the laminated structure of bulletproof materials, the interlaminar region is usually a weak area in terms of mechanical properties. Under external impact or long-term service loads, microscopic damage mechanisms such as interface slip, microcrack initiation and propagation, and fiber-matrix debonding may occur. These microscopic processes are all accompanied by the irreversible conversion of mechanical energy into thermal energy, i.e., entropy production.

[0018] In this embodiment, the interlayer entropy yield is not obtained directly through temperature measurement, but indirectly extracted by analyzing the energy dissipation characteristics of the ultrasonic interferometric signal. When a shear wave propagates in a damaged interlayer region, the wave energy attenuates due to mechanisms such as interfacial friction and the mutual displacement of crack surfaces, and the degree of attenuation is related to the damage evolution rate. This energy dissipation process follows the second law of thermodynamics, and its dissipation power has a definite physical relationship with the entropy yield.

[0019] By performing time-frequency decomposition on the received signal, energy components with different propagation paths and polarization states can be separated, thereby identifying abnormal energy dissipation regions caused by interlayer damage. Therefore, the interlayer entropy yield becomes a bridge connecting ultrasonic detection signals and the internal damage state of materials, and can sensitively reflect the early weak interlayer bonding state before macroscopic stratification has formed.

[0020] In some embodiments, interlaminar damage is defined as a continuous intrinsic variable to describe the gradual process of interlaminar interface separation from intact to completely separated. This variable can be mathematically characterized by an interlaminar phase field order parameter, the range of which corresponds to the continuous transition from the state of material integrity to complete delamination.

[0021] There is a thermodynamic conjugate relationship between interlayer entropy production and interlayer damage degree. Specifically, damage evolution is a non-equilibrium thermodynamic process. The product of the generalized force driving damage development (related to damage potential energy) and the damage evolution rate determines the entropy production per unit volume. This conjugate relationship reflects the essential link between energy dissipation and structural degradation: the faster the damage evolution, the more intense the energy dissipation, and the higher the entropy production; conversely, by monitoring the spatial distribution of entropy production, the activity level of damage evolution can be deduced.

[0022] During the detection process, a bijective mapping relationship between the interlayer phase field distribution and the entropy yield field is established. Hard-constraint encoding using a reversible neural network structure ensures that the two strictly satisfy the thermodynamic conjugate relationship. This constraint not only guarantees the dimensional consistency of physical quantities but also ensures that the detection results conform to the fundamental laws of thermodynamics, avoiding non-physical results that might arise from purely data-driven methods.

[0023] S3, based on the interlayer phase field distribution and the fiber orientation gradient of the ballistic material to be tested, uses a damage evolution model that includes a geometrically necessary orientation gradient dissipation term to deduce the predicted spatial distribution of the interlayer phase field at the next moment. It needs to be explained that the interlaminar damage evolution of bulletproof materials depends not only on the local stress state, but also on the significant influence of the spatial non-uniformity of fiber orientation. In laminated structures, the fiber orientations of adjacent plies are often staggered at specific angles (such as 0°, 90°, ±45°), and this orientation variation forms a geometrically necessary orientation gradient at the interlaminar interface.

[0024] The geometrically necessary orientation gradient dissipation term characterizes the additional energy dissipation mechanism caused by spatial variations in fiber orientation. When interlaminar shear deformation occurs, the sudden change in fiber orientation leads to a complex distribution of the interlaminar stress field, generating additional geometrically necessary dislocations or orientation mismatches, resulting in viscous dissipation. This dissipation effect is mathematically represented by the dynamic viscosity parameter, which is not a material constant but is related to the magnitude of the orientation gradient.

[0025] Introducing this term into the damage evolution model enables the model to accurately describe the interlaminar damage evolution characteristics in regions with complex fiber orientations (such as curvature changes in helmets or edge reinforcement areas of inserts). Regions with high orientation gradients often exhibit higher apparent viscosity, resulting in a correspondingly slower damage propagation rate. This physical mechanism is quantitatively reflected in the prediction model through the geometrically necessary orientation gradient dissipation term.

[0026] S4, adjust the excitation phase difference and ultrasonic intensity of the shear wave in real time according to the spatial gradient of the predicted spatial distribution, and repeat S1 to S3 based on the adjusted excitation phase difference and ultrasonic intensity.

[0027] This embodiment of the method employs a closed-loop adaptive strategy, adjusting the excitation parameters of the shear wave in real time based on the predicted interlayer phase field distribution. The optimization objective of this closed-loop system is to improve detection sensitivity and spatial resolution for potentially damaged areas, and to reduce ultrasonic energy injection to avoid unnecessary material excitation for intact areas.

[0028] In practice, the system assesses the evolution trend of the damage front based on the spatial gradient of the interlayer phase field distribution at the current moment. When a significant change in damage gradient is predicted to occur in a certain region at the next moment, the system automatically adjusts the phase difference between the two shear waves to maximize the sensitivity of the interference field in that region, while appropriately increasing the ultrasonic intensity to enhance the signal-to-noise ratio. Conversely, in homogeneous and intact regions, the excitation intensity is maintained or reduced.

[0029] The iterative process is implemented through a model predictive control framework. The termination condition can be set to reach a preset detection accuracy threshold, or to stop when the spatial distribution change of entropy yield in two consecutive iterations is less than the convergence tolerance. This adaptive mechanism ensures that the detection process can detect potential damage without placing an excessive detection burden on in-service equipment.

[0030] Further, S2, the interference signal is decomposed into time and frequency components to extract the spatial distribution of interlayer entropy yield, and the interlayer phase field distribution is calculated and determined based on the thermodynamic conjugate relationship between interlayer entropy yield and interlayer damage degree, including: Construct the time-varying Stokes parameters of the interference signal and build the polarization coherence tensor; To accurately extract the interlayer entropy yield, polarization state analysis of the acquired interferometric signals is required. This implementation constructs time-varying Stokes parameters to comprehensively describe the polarization characteristics of the received wavefield. Signal components with different polarization directions are acquired through multiple receiving channels, and the four elements of the Stokes parameters are calculated, corresponding to the total horizontal light intensity (or sound intensity), the difference between the horizontal and vertical polarization components, the difference between the polarization components at 45 degrees, and the circular polarization component.

[0031] Based on these parameters, a polarization coherence tensor is constructed. This tensor is a second-order complex matrix whose elements reflect the statistical correlation between different polarization states. This tensor comprehensively characterizes the polarization coherence properties of the wave field during time-varying processes, including mixed information from fast and slow shear wave modes. By analyzing the eigenvalues ​​of this tensor, the influence of material anisotropy on wave propagation can be identified.

[0032] The polarization coherence tensor is decomposed eigenvalues ​​to obtain the first eigenpolarization state corresponding to the fast shear wave mode and the second eigenpolarization state corresponding to the slow shear wave mode. The interference signal is then projected onto the first and second eigenpolarization states to achieve polarization whitening separation. The fast shear wave mode (qS1) refers to the shear wave component with a high phase velocity propagating along the principal direction of the bulletproof material fibers under test; the slow shear wave mode (qS2) refers to the shear wave component with a low phase velocity propagating perpendicular to the fiber direction. The separation between the two originates from the intrinsic modulation of the elastic stiffness tensor of the anisotropic laminate to the propagation velocity of shear waves in different polarization directions, determined by two different eigenvalues ​​of the Christoffel equation. Among these, the slow shear wave mode (the second intrinsic polarization state) is more sensitive to interlaminar shear stiffness degradation and is therefore chosen for entropy yield extraction.

[0033] Polarization whitening separation aims to decompose mixed interference signals into independent fast and slow shear wave components. Due to the significant anisotropy in the interlayer regions of bulletproof materials, shear waves propagating within them split into two modes with orthogonal polarization directions and different propagation speeds. These two modes superimpose at the receiver, forming a complex interference pattern.

[0034] By performing eigenvalue decomposition on the polarization coherence tensor, two eigenstates are obtained, corresponding to the principal polarization directions of the fast and slow waves, respectively. Projecting the original interference signal onto these two eigenstates allows for the separation of the two modes. "Whitening" here means decorrelation of the two separated signal components, i.e., eliminating their statistical correlation, making their respective energy distributions independent of the other component, thus facilitating the separate analysis of the energy attenuation characteristics of each mode.

[0035] Synchronous squeezing transformation is performed on the first and second intrinsic polarization states after separation. The spatial distribution of interlayer entropy yield is determined according to the instantaneous energy decay characteristics of each polarization state. The redistribution bandwidth parameter is adjusted according to the interlayer entropy yield. The redistribution bandwidth parameter is inversely proportional to the interlayer entropy yield. The higher the entropy yield, the stronger the time-frequency clustering. For the two separated intrinsic polarization state signals, synchronous squeezing transform is used for time-frequency analysis. Compared with traditional short-time Fourier transform or wavelet transform, synchronous squeezing transform has higher time-frequency convergence and can more accurately identify the instantaneous frequency components in the signal and their energy changes over time.

[0036] In this embodiment, synchronous extrusion transform is particularly suitable for capturing the instantaneous energy attenuation characteristics of shear waves propagating in interlaminar damage regions. Since energy dissipation caused by damage often manifests as rapid attenuation within a specific time period, synchronous extrusion transform uses a redistribution technique to compress the ambiguous energy distribution to near the true instantaneous frequency, thereby clearly distinguishing energy dissipation events caused by mechanisms such as interlaminar friction and microcrack propagation.

[0037] To further improve the resolution of time-frequency analysis, this implementation introduces an adaptive redistribution bandwidth parameter. This parameter controls the degree of localization of frequency redistribution in synchronous extrusion transform. In high-entropy yield regions, i.e., regions where interlayer damage is more active and energy dissipation is more severe, the redistribution bandwidth is automatically reduced to enhance time-frequency clustering, thereby more finely distinguishing the energy decay characteristics caused by damage; in low-entropy yield regions, the bandwidth is appropriately widened to ensure the stability of the analysis.

[0038] This adaptive logic is based on the physical relationship between entropy yield and damage activity: a higher entropy yield indicates more intense damage evolution and richer non-stationary features in the signal, requiring higher time-frequency resolution for capture; conversely, in stable damage regions, a coarser analysis granularity can be used. By dynamically adjusting the bandwidth parameter, an optimal balance between computational resources and detection accuracy is achieved.

[0039] Based on the thermodynamic conjugate relationship between interlayer entropy yield and interlayer damage, an initial estimate of the interlayer phase field distribution is calculated and determined.

[0040] It should be explained that interlaminar damage is a physical quantity that describes the degree of degradation of the mechanical properties of the interlaminar interface, characterizing the continuous transition of a material from an intact state to a completely delaminated state. It reflects the macroscopic cumulative effect of microscopic damage mechanisms such as fiber-matrix interface debonding and microcrack initiation and propagation.

[0041] The interlayer phase field order parameter is a continuous field mathematical description of the damage degree mentioned above. It transforms the discrete damage state into a continuously distributed field variable in space through phase field theory. This order parameter is usually normalized, for example, it takes a value of zero at the intact interface, approaches a saturation value (such as 1) in the fully delaminated region, and intermediate values ​​represent the transition state of partial damage.

[0042] The specific relationship between the two is reflected in the fact that the spatial distribution of the interlayer phase field order parameter is the spatial distribution of the interlayer damage degree, and the magnitude of the order parameter directly quantifies the severity of damage at the corresponding location.

[0043] By solving the thermodynamic conjugate relationship between entropy production rate and interlayer damage degree, we actually obtain the initial distribution of the interlayer phase field order parameter. The damage evolution equation also uses this order parameter as the state variable to deduce its temporal evolution, that is, to predict the development of interlayer damage degree. The damage evolution potential energy density defines the generalized force per unit volume with respect to the rate of change of the interlayer phase field order parameter. This shows that the change of the order parameter is directly related to the energy cost required for the change of damage degree. Calculating and determining the initial estimate of the interlayer phase field distribution is essentially determining the initial spatial distribution of interlayer damage degree.

[0044] Based on the embodiments provided in this application, polarization whitening separation is achieved by constructing time-varying Stokes parameters of the interference signal and building a polarization coherence tensor, followed by eigenvalue decomposition. This process effectively isolates the fast and slow shear wave modes directly related to material anisotropy. On this basis, synchronous squeezing transformation is performed on the separated polarization states, and interlayer entropy yield is extracted based on instantaneous energy decay characteristics. This design cleverly utilizes the differences in energy dissipation modes of different polarization states. In particular, by establishing an inverse relationship between the redistribution bandwidth parameter and the interlayer entropy yield, time-frequency analysis can adaptively focus on high-entropy-yield (i.e., high dissipation or potential damage) regions, thereby significantly enhancing the convergence of time-frequency characterization and the ability to identify early damage characteristics in complex signal backgrounds, providing a cleaner and more sensitive data foundation for subsequent phase-field inversion.

[0045] Furthermore, determining the spatial distribution of interlayer entropy yield includes: The energy decay rate is calculated based on the energy decay characteristics of the second intrinsic polarization state within the time window. When calculating the interlayer entropy yield, a time window analysis is required to examine the energy changes of the slow shear wave mode (second intrinsic polarization state). The length of the time window must be chosen by comprehensively considering both the propagation time of the ultrasound in the material and the physical timescale of the energy decay process.

[0046] Specifically, the window start point is set at the initial moment when the shear wave enters the interlayer region to be detected. The window length should include at least several shear wave cycles to ensure that a stable energy attenuation trend can be observed, while not being too long to avoid including too many reflected signals from other interfaces. For bulletproof laminates of typical thickness, the window length is usually set to 1.5 to 2 times the propagation time of the shear wave in the interlayer to ensure coverage of the main energy dissipation processes while avoiding boundary effect interference.

[0047] Calculate the cumulative energy of the second intrinsic polarization state within the time window. Regarding time The energy decay rate is obtained by differentiation. ; Energy decay rate Divide by the absolute temperature within the time window With effective volume The product of these factors yields the interlayer entropy yield. ; .

[0048] The calculation of interlayer entropy production involves the absolute temperature of the detection area. In this embodiment, non-destructive temperature measurement methods are preferred to obtain the actual temperature. The surface temperature of the area to be detected can be scanned by an infrared thermal imager, and the internal temperature distribution of the interlayer can be estimated by combining it with a heat conduction model; or an embedded fiber optic temperature sensor (if the equipment under test allows for pre-embedding) can be used to directly obtain the temperature profile in the depth direction.

[0049] For situations where the internal temperature cannot be directly measured, the ambient temperature can be used as an approximation, but it needs to be corrected based on the ultrasonic energy injection during the detection process. Considering that the ultrasonic intensity used in this detection method is relatively low, the resulting temperature rise is usually negligible. Therefore, using the ambient temperature superimposed with an empirical correction value is sufficient to meet the accuracy requirements for entropy yield calculation.

[0050] Effective volume refers to the spatial scale of the material involved in the entropy production process, that is, the volume of the region where ultrasound interacts with interlaminar damage and generates irreversible energy dissipation. This volume is related to the waist radius of the ultrasound beam, the thickness of the interlaminar interface, and the lateral distribution of the damaged region.

[0051] In practical testing, the effective volume can be estimated through the spatial distribution characteristics of the ultrasonic beam. Based on the transducer aperture, operating frequency, and sound velocity in the material, the focusing or diffusion volume of the ultrasonic wave in the interlayer region is calculated. For delamination damage, the effective volume is approximately the product of the projected area of ​​the damaged region in the interlayer plane and the interlayer thickness. By establishing an ultrasonic propagation model and combining it with the spatial distribution characteristics of the array-received signal, the effective volume of the current detection point can be derived, thereby normalizing the energy attenuation rate to the entropy yield per unit volume, ensuring the comparability of results from different detection locations and damage scales.

[0052] Based on the embodiments provided in this application, the energy decay rate of slow shear wave modes (which are generally more sensitive to interlaminar states) within a time window is derived by combining absolute temperature and effective volume. This calculation formula has clear physical meaning, directly linking the measurable signal energy decay to the concept of entropy production in thermodynamics, providing a concrete and operable implementation for assessing irreversible energy dissipation (i.e., damage or plastic deformation processes) within materials. This enables the detection method to transform the abstract concept of damage into a computable scalar field based on energy observations.

[0053] Furthermore, S2 calculates and determines the interlayer phase field distribution, including: Construct a fiber bundle network diagram, where the node positions correspond to the geometric centers of each fiber bundle; the weight of the edge connecting adjacent nodes in the fiber bundle network diagram is the product of the interlaminar shear stiffness between adjacent nodes and the square of the cosine of the angle between the fiber orientations of adjacent nodes. Based on the fiber bundle network diagram, the frequency domain Green's function of the simplified Christoffel equation, which retains only the dominant term of interlaminar shear stiffness, is solved to obtain the rough wave propagation kernel; To accurately describe the mechanical transmission path within the bulletproof material, this embodiment constructs a fiber bundle network diagram to characterize the interlayer structural features of the material. The construction of this network diagram is based on the digital characterization of the material's microstructure: first, spatial distribution images of the fiber bundles within the material are obtained through micro-CT scanning or high-resolution ultrasound imaging; the geometric center positions of each fiber bundle are identified; and these center points are mapped to nodes in the network diagram.

[0054] The connection relationships between nodes are determined based on the spatial adjacency of fiber bundles: if two fiber bundles have mechanical interactions at the interlayer interface (i.e., the spatial distance is less than several times the characteristic diameter of the fiber bundles), then an edge is established between the corresponding nodes. The weighting of the edges must consider both mechanical properties and geometric characteristics: the interlayer shear stiffness between adjacent fiber bundles is used as the basic mechanical parameter, multiplied by the square of the cosine of the angle between the two fiber bundle orientations as a geometric correction factor. This design reflects the influence of fiber bundle orientation differences on interlayer load transfer efficiency; that is, when the two fiber bundles are perpendicular, the cosine value is zero, the edge weight is minimal, indicating the weakest interlayer shear transfer; when the orientations are parallel, the weight reaches its maximum. Through this weighting method, the network diagram can accurately capture the non-uniformity of interlayer mechanical properties caused by fiber layup interleaving.

[0055] It should be noted that the Christoffel equation is a classical governing equation in elasticity that describes the propagation characteristics of waves in anisotropic media. Its establishment and the method for solving the frequency domain Green's function are existing technologies in this field, and will not be elaborated here.

[0056] The simplified Christoffel equation in this application refers to an approximation within the classical equation framework, taking into account the physical property that the interlaminar shear stiffness of bulletproof materials is much lower than the in-plane stiffness (typically 2-3 orders of magnitude lower). It neglects higher-order minor quantities of bending stiffness and in-plane tensile stiffness, retaining only the dominant term of interlaminar shear stiffness. The rough wave propagation kernel is a frequency-domain Green's function obtained by solving this simplified equation. It characterizes the idealized wave propagation characteristics after neglecting microscopic defects such as fiber ripples and local debonding, serving as the physical prior for subsequent neural operator learning.

[0057] Specifically, when solving for wave propagation characteristics, the traditional Christoffel equation requires consideration of the complete elastic constant tensor of the material, involving numerous elastic modulus parameters. However, in this detection scenario, the focus is on the shear characteristics of the interlaminar interface, and interlaminar damage mainly manifests as degradation of shear stiffness. Therefore, this embodiment adopts a simplification strategy: in the frequency domain Green's function solution, only the dominant term of interlaminar shear stiffness is retained, while the explicit consideration of secondary coupling terms such as compressive modulus is ignored.

[0058] The reason for this simplification is that, for laminated bulletproof materials, the mechanical weakness of the interlaminar region lies in its low interlaminar shear stiffness, a parameter that is most sensitive to damage conditions. Furthermore, the in-plane modulus is typically two orders of magnitude higher than the interlaminar shear modulus, and its contribution to the interlaminar response in shear wave propagation problems can be considered as a background field. This simplification significantly reduces computational complexity, transforming the original full matrix equations into quickly solvable scalar or low-order matrix equations, thus meeting the computational efficiency requirements of real-time online detection.

[0059] The resulting frequency-domain Green's function is called the rough wave propagation kernel, which provides a baseline physical picture of shear wave propagation in the interlayer region. Although this kernel neglects some minor effects, it accurately captures the dominant physical mechanism of interlayer shear wave propagation, providing a reasonable initial estimate for subsequent fine-tuning.

[0060] A Fourier neural operator is constructed, and the Fourier transform of the rough wave propagation kernel is used as the initial convolution kernel of the neural operator. The learning objective of the neural operator is the residual mapping between the real interference signal and the rough wave propagation prediction. To bridge the gap between the coarse wave propagation kernel and the real complex propagation behavior, this embodiment introduces a Fourier neural operator for hybrid modeling. This architecture employs a strategy combining a physically initialized kernel with data-driven residual learning: first, the coarse wave propagation kernel is Fourier transformed and used as the initial convolution kernel for the neural operator, meaning the network already possesses basic physical propagation characteristics during the initialization phase.

[0061] The learning objective of the neural operator is not to directly fit the complete wavefield, but rather to learn the residual mapping between the real interference signal and the rough wave propagation prediction. This design fully utilizes the physical model's ability to describe the main laws and the neural network's ability to fit complex nonlinear effects (such as material viscoelasticity, incomplete interface contact, edge diffraction, etc.). The network corrects the propagation kernel in the frequency domain through convolution operations, maintaining physical interpretability (frequency domain correction corresponds to physical correction of wave velocity and attenuation) while possessing sufficient flexibility to adapt to the complex characteristics of real materials.

[0062] Construct a reversible neural network layer, which includes an affine coupling structure, and construct a bijective mapping pair between the interlayer phase field order parameter and the interlayer entropy yield. To establish a thermodynamically consistent mapping between interlayer phase field order parameters and interlayer entropy yield, this embodiment employs a reversible neural network structure. This network is built upon an affine coupling structure and achieves a precise bidirectional mapping between the phase field distribution and the entropy yield field through reversible transformations, ensuring no information loss and that the mapping is reversible.

[0063] The key lies in the hard constraint encoding of the thermodynamic conjugate relationship. A pre-defined multiplication node is set in the scaling branch of the affine coupling structure. This node is not a learnable weight parameter, but rather a enforced physical computation unit. Specifically, this node multiplies the generalized force per unit volume by the evolution rate of the interlayer phase field order parameter, divides by the absolute temperature, and forcibly assigns the result as the interlayer entropy yield. This constraint is rigidly embedded during the network's forward propagation phase, rather than being achieved through backpropagation optimization of the loss function.

[0064] This hard constraint differs from the soft constraint approach of traditional physical information neural networks (which forces the network to approximately satisfy physical relationships through a penalty term in the loss function). The hard constraint ensures that the thermodynamic conjugate relationship is strictly adhered to regardless of how the network parameters are updated, thus guaranteeing that the inversion result necessarily satisfies the second law of thermodynamics and avoiding non-physical predictions. Through this design, the network synchronously outputs a physically consistent interlayer phase field distribution and its rate of change over time during the inversion process.

[0065] In the multiplication branch of the affine coupling structure, a preset multiplication operation node is set to represent the generalized force per unit volume. Evolution rate of interlayer phase field order parameters Multiply and then divide by absolute temperature The result of this calculation is constrained to be equal to the interlayer entropy yield. Thermodynamic conjugate relationships are achieved by pre-setting hard constraint encoding of multiplication operation nodes. ; The interlayer phase field distribution and its temporal rate of change are determined by synchronous inversion using hard-constrained coding, rather than by approximating the thermodynamic conjugate relationship through loss function optimization.

[0066] Based on the embodiments provided in this application, a coarse wave propagation physical kernel is obtained by constructing a fiber bundle network diagram and solving a simplified Christoffel equation. This kernel is then used to initialize the Fourier neural operator, ensuring that the model has a good starting point for physical interpretability. Specifically, a reversible neural network layer including an affine coupling structure is introduced, and the thermodynamic conjugate relationship is hard-constrained and encoded through preset multiplication operation nodes. This design ensures that the neural network must strictly satisfy the physical law that "the entropy production rate equals the product of the generalized force and the phase field evolution rate divided by the temperature" during training and inference. This strongly guarantees that the inverted interlayer phase field distribution and its evolution rate are thermodynamically self-consistent and reliable, effectively avoiding the physically unreasonable output that may occur in purely data-driven methods.

[0067] Furthermore, the generalized force per unit volume The damage evolution potential energy density is obtained by taking the partial derivative of the interlayer phase field order parameter. This density includes a coupling term between the characteristic length scale of the interlayer crack and the critical strain energy release rate, resulting in a generalized force per unit volume. It has the dimensions of energy / volume.

[0068] The generalized force per unit volume is obtained by differentiating the damage evolution potential energy density with respect to the interlaminar phase field order parameter. The design of this potential energy density function needs to reflect the physical characteristics of interlaminar cracks: on the one hand, it includes the characteristic length scale parameter of interlaminar cracks, which determines the spatial range of damage localization and prevents grid dependence in numerical calculations; on the other hand, it includes the critical strain energy release rate parameter, which characterizes the inherent toughness of interlaminar materials against crack propagation.

[0069] In the potential energy density function, these two parameters appear in a coupled form: the characteristic length scale of the interlaminar crack controls the gradient energy term of the phase field variable, smoothing the damage transition region; while the critical strain energy release rate controls the energy threshold of damage evolution. When the phase field order parameter changes, the rate of change of the potential energy density automatically reflects the combined effect of these two parameters—the smaller the crack size and the lower the toughness, the more drastic the energy change caused by the phase field change.

[0070] The generalized force per unit volume defined in the above manner has the dimension of energy per unit volume. This dimensional characteristic ensures self-consistency within the entire thermodynamic framework: the product of the generalized force and the phase field evolution rate has the dimension of energy per unit volume per unit time (i.e., power density), which, when divided by temperature, yields the entropy yield (entropy per unit volume per unit time), perfectly consistent with thermodynamic theory. Based on the embodiments provided in this application, the generalized force per unit volume... The potential energy density is derived from the phase field order parameter by differentiating the damage evolution potential energy density, which includes a coupling term between the characteristic length of interlaminar cracks and the critical strain energy release rate. The entire phase field inversion and subsequent evolution model is built on a solid foundation of continuum mechanics and thermodynamics, improving the physical realism of the model predictions and its applicability to problems of different materials and scales.

[0071] Furthermore, S3, based on the interlayer phase field distribution and the fiber orientation gradient of the ballistic material to be tested, uses a damage evolution model including a geometrically necessary orientation gradient dissipation term to deduce the predicted spatial distribution of the interlayer phase field distribution at the next time step, including: The three-dimensional fiber orientation field of the ballistic material under test can be obtained by polarization-sensitive optical coherence tomography or by shear wave birefringence inversion. ; Fiber orientation fields are fundamental to describing the anisotropic characteristics of materials. This embodiment provides two technical approaches for obtaining three-dimensional fiber orientation fields, both of which are well-known technologies and will not be elaborated upon here: Firstly, polarization-sensitive optical coherence tomography (OCT). This technique utilizes the birefringence of near-infrared light in fiber materials to reconstruct the optical axis orientation of the fiber bundle by measuring the polarization state change of the backscattered light. This technique offers micrometer-level resolution, is suitable for high-precision characterization in laboratory environments, and can obtain the complete three-dimensional orientation distribution of helmets or plates.

[0072] Secondly, the shear wave birefringence inversion method is used. This method utilizes a pre-arranged transducer array to transmit and receive shear waves with different polarization directions, measuring the velocity difference and polarization direction between fast and slow waves. Based on the theory of elastic wave propagation in anisotropic media, the dominant fiber orientation is deduced. This method is suitable for in-service equipment inspection, requiring no disassembly or special pretreatment, but its resolution is relatively low; it can be used as a rapid initial screening method.

[0073] Calculate the Nye curvature tensor of the three-dimensional fiber orientation field The geometrically necessary orientation gradient magnitude is obtained. , For the gradient operator, the Nye curvature tensor This constitutes a geometrically necessary orientation gradient; After obtaining the orientation field, its spatial curl is calculated to obtain the Nye curvature tensor. The physical meaning of this tensor lies in quantifying the degree of local bending or twisting of fiber orientation: when the fiber bundle is aligned straight in space, the curvature tensor is zero; when the fiber bundle bends or there is an abrupt change in orientation between adjacent layers, the curvature tensor exhibits a non-zero value. Taking its modulus yields the magnitude of the geometrically necessary orientation gradient. The larger this value, the more drastic the spatial change in fiber orientation, corresponding to the curvature change region of the helmet or the edge transition region of the insert.

[0074] Based on the geometrically necessary orientation gradient magnitude Calculate discrete ripple energy storage density and according to The modified interlayer critical strain energy release rate was obtained Discrete ripple energy storage density With interlaminar critical strain energy release rate This constitutes an energy competition term; The discrete ripple energy storage density is propagated below. This calculation method is merely another embodiment of the present invention and is not the only one. The calculation of the discrete ripple energy storage density does not affect the implementation of the above technical solution.

[0075] ,in Interlaminar shear modulus The characteristic diameter of the fiber bundle; The existence of a geometrically necessary orientation gradient implies the presence of geometrically necessary dislocations or orientation mismatches in the material's microstructure. This microscopic inconsistency generates additional elastic energy storage when subjected to shear loads between layers, termed discrete ripple energy storage density. This energy term corresponds to the elastic strain energy stored in the interlayer interface region due to spatial variations in fiber orientation, and its magnitude is proportional to the square of the orientation gradient modulus.

[0076] Fiber orientation gradient significantly affects the propagation resistance of interlaminar cracks. In regions with high orientation gradients, crack propagation requires overcoming additional geometric obstacles and energy dissipation, manifested as an increase in the effective critical strain energy release rate. Therefore, this embodiment corrects the nominal critical strain energy release rate based on the orientation gradient modulus: the larger the orientation gradient, the higher the corrected critical value, indicating greater difficulty in damage propagation. This correction reflects the true constraint of material microstructure on damage evolution.

[0077] Based on the evolution rate equation of interlayer phase field sequence parameters Computational evolution, in which, Represents the driving force per unit volume. Discrete ripple energy storage density Determined together with the interlayer phase field distribution; The damage evolution threshold, Indicates the evolution rate of interlayer phase field order parameters; dynamic viscosity This is a geometrically necessary orientation gradient dissipation term, used to characterize interlayer viscous dissipation caused by fiber orientation gradient; driving force per unit volume The dimension of is energy / volume. The damage evolution threshold and its dimensions are... same, The dynamic viscosity is expressed in the form of energy·time / volume; Based on the evolution rate and the current time step, calculate the predicted spatial distribution of the interlayer phase field at the next time step.

[0078] The evolution of the interlayer phase field order parameter follows a rate equation, in which each variable has a definite physical role: The driving force per unit volume characterizes the net energy release rate that propels damage development, and is jointly determined by the discrete corrugated energy storage density and the current damage state. The damage evolution threshold is the inherent resistance of the material to damage initiation; when the driving force is below this threshold, damage does not develop; damage only begins to evolve when the driving force exceeds the threshold. This threshold can be determined through standard interlaminar fracture tests (such as double cantilever beam tests or end-notch bending tests). For typical aramid fiber laminates, the value is usually in the range of 0.1 to 1.0 kJ / m², with the specific value depending on the toughness of the resin matrix and the fiber surface treatment process.

[0079] Dynamic viscosity characterizes the energy dissipation properties during damage evolution, particularly the interlaminar viscous dissipation caused by fiber orientation gradients. This parameter was calibrated by inversion from V50 ballistic limit test data and corrected online based on real-time detection data using a Kalman filter algorithm.

[0080] Macaulay brackets represent the threshold characteristics of damage evolution. When the difference between the driving force and the threshold is positive, the brackets contain that difference, and the damage evolves at a positive rate. When the difference is zero or negative, the brackets contain zero, and the damage rate remains zero, meaning the damage neither expands nor heals. This mathematical treatment rigorously characterizes the irreversibility and threshold characteristics of damage evolution. Based on the calculated evolution rate and the current time step, the interlayer phase field distribution at the next moment can be deduced, enabling the prediction of damage development trends.

[0081] Based on the embodiments provided in this application, the degree of local bending or twisting of fiber arrangement is quantified by calculating the Nye curvature tensor of the fiber orientation field. It should be explained that in fiber-reinforced composite materials such as bulletproof materials, the drastic changes in fiber orientation (high curvature) themselves store elastic strain energy (discrete ripple energy storage), which competes with the propagation of interlaminar damage. This gradient term and the critical strain energy release rate modified by it are incorporated into the damage evolution model, and the associated dynamic viscosity is defined. As a dissipative term, it enables the model to more realistically reflect the constraints and guiding effects of fiber architecture on the path of damage initiation and propagation, thereby significantly improving the accuracy of damage evolution trend prediction under complex fiber orientation distribution.

[0082] Furthermore, driving force per unit volume The calculations include: Calculation of discrete ripple energy storage density The resulting elastic restoring force, the elastic restoring force through the geometrically necessary orientation gradient magnitude The product of the fiber bundle characteristic diameter is used to determine the energy / volume dimension; The elastic restoring force in the unit volume driving force originates from the release tendency of the discrete corrugated energy storage density. In regions where fiber orientation changes spatially, the material stores additional elastic strain energy generated by the geometrically necessary orientation gradient. This energy storage state is thermodynamically non-equilibrium and tends to revert to a lower energy state. When damage evolves between layers, the microscopic inconsistencies caused by the fiber orientation gradient produce a restoring effect, attempting to homogenize the interlaminar deformation. The magnitude of this restoring force directly depends on the intensity of the geometrically necessary orientation gradient and the characteristic dimensions of the fiber bundles, reflecting the geometrical resistance of the material's microstructure to damage propagation.

[0083] Calculate the dissipative force caused by the entropy productivity gradient, with the dimension being energy / volume; The dissipation force originates from the spatial inhomogeneity of entropy production rate. During damage evolution, the energy dissipation rate varies in different regions, forming an entropy production rate gradient. This inhomogeneity drives the damage state to shift towards regions with more sufficient dissipation, manifesting as a dissipation mechanism that promotes a more uniform damage distribution. The existence of this force reflects the constraint of the second law of thermodynamics on the direction of damage evolution, that is, the system tends to optimize the energy dissipation path by adjusting the damage distribution.

[0084] The elastic restoring force and the dissipated force are superimposed to obtain the driving force G per unit volume.

[0085] The superposition of elastic restoring force and dissipative force constitutes the total driving force per unit volume, reflecting the competition mechanism between energy storage and energy dissipation during damage evolution. Elastic restoring force represents the elastic response of the material's microstructure, attempting to maintain structural integrity; dissipative force represents the driving force of the thermodynamically irreversible process, promoting damage evolution. The superposition of these two forces determines the net trend of damage evolution: when elastic restoring force dominates, damage propagation is inhibited; when dissipative force dominates, damage development accelerates. This superposition logic ensures that the damage evolution model simultaneously satisfies the principles of mechanical equilibrium and the fundamental laws of thermodynamics.

[0086] Based on the embodiments provided in this application, the driving force G per unit volume is decomposed into two parts: the elastic restoring force caused by the geometrically necessary orientation gradient and the dissipation force caused by the entropy yield gradient. This decomposition reveals two different physical mechanisms driving damage evolution: the former originates from the redistribution of internal stress due to the inhomogeneity of the material's microstructure, while the latter is related to the inhomogeneity of the energy dissipation process. By superimposing these two forces to comprehensively determine the driving force, the evolution model can simultaneously capture the combined influence of the material's microstructure characteristics and the current damage state on subsequent evolution, thereby making the predicted damage propagation direction and rate more consistent with the actual physical response of the material.

[0087] Furthermore, the dynamic viscosity μ is determined by calibration using the following steps: V50 ballistic limit test was performed on standard specimens to obtain the interlaminar delamination area under different impact velocities; The V50 ballistic limit test is a standardized test method for evaluating the interlaminar impact resistance performance in the field of ballistic materials. It is an existing technology in this field, and its specific implementation steps will not be described in detail here.

[0088] The innovation of this application lies in establishing a quantitative correspondence between the macroscopic interlayer delamination area obtained from the V50 test and the interlayer phase field sequence parameter, including: matching the measured delamination area under different impact velocities with the phase field distribution predicted based on the damage evolution model, and using Bayesian inversion or optimization algorithms to determine the initial value of dynamic viscosity μ, thereby realizing the cross-scale correlation between macroscopic ballistic performance and microscopic thermodynamic state parameters.

[0089] The dynamic viscosity parameters need to be initially calibrated through standardized tests. The V50 ballistic limit test is used as the calibration basis. This test determines the critical velocity at which the probability of penetration is 50% by launching standard fragments that impact material samples at different velocities. After the test, the samples are dissected or ultrasonically scanned to accurately measure the macroscopic area of ​​interlayer delamination.

[0090] A correlation was established between the macroscopic delamination area and the interlayer phase field order parameter: the delamination region was defined as the area where the phase field order parameter reached a critical value, and the equivalent damage volume corresponding to the phase field distribution was calculated by integration. The measured delamination area under different impact velocities was compared with the damage distribution predicted based on the current viscosity parameter, and the viscosity value was adjusted to make the prediction results match the measured data. When the initial viscosity value minimized the statistical deviation between the predicted delamination area and the measured value within the typical impact velocity range, it was determined as the initial dynamic viscosity of this batch of materials.

[0091] The initial value of dynamic viscosity μ is determined by inversion based on the integral correspondence between the interlayer phase field sequence parameter and the interlayer layer area. The dynamic viscosity μ is corrected online based on real-time detection data using the Kalman filter algorithm.

[0092] like Figure 2 As shown, in one specific embodiment, the core model used in the damage evolution prediction step of this invention is a physical model coupling fiber orientation gradient with damage evolution. This model aims to establish and quantify the physical relationship between the microscopic fiber structure of the material and macroscopic damage behavior, thereby improving the accuracy of the prediction.

[0093] The model's construction begins with a mathematical description of the material's anisotropic microstructure. A three-dimensional fiber orientation field, obtained through techniques such as polarization-sensitive optical coherence tomography, fully records the fiber alignment at every point within the material. Performing curl calculations on this orientation field yields a physical quantity called the Nye curvature tensor, which quantitatively characterizes the degree of local bending and twisting of the fiber orientation in space—the fiber orientation gradient. This non-uniform fiber bending is not dissipationless; it introduces and stores a portion of elastic strain energy in the interlayer interface region, defined as discrete ripple energy storage.

[0094] This model embeds the aforementioned microscopic physical mechanisms into a macroscopic framework of damage evolution dynamics. In the model, damage development is controlled by the evolution equation of the interlayer phase field order parameter. In this equation, the generalized force driving damage propagation consists of two parts: one part originates from the elastic recovery effect corresponding to the discrete corrugated energy storage; the other part is related to the non-uniform distribution of energy dissipation (entropy production) within the system. The critical threshold for the material's resistance to damage propagation is directly related to a modified interlayer critical strain energy release rate, which has been shown to be proportional to the magnitude of the fiber orientation gradient. This means that regions with more drastic fiber orientation changes and more complex textures have a stronger inherent ability to suppress damage initiation and resist crack propagation. Simultaneously, the equivalent viscosity coefficient in the model is defined as a dissipation term related to the orientation gradient, describing the viscous drag effect of the fiber network during the damage process.

[0095] Therefore, this coupled model reveals a clear energy competition and allocation mechanism: energy input to the material system (e.g., impact kinetic energy or ultrasonic excitation energy) is partly converted into discrete corrugated energy storage that causes bending deformation of the fiber network, and partly used to drive the evolution and expansion of interlaminar damage (phase field). The greater the fiber orientation gradient, the stronger the tendency for energy to be converted into corrugated energy storage, thus consuming more energy available for damage expansion. Through this physical principle-based modeling, the system can achieve more reliable predictions of the anisotropic damage evolution behavior of ballistic materials.

[0096] Since the material condition of actual in-service equipment may change due to factors such as aging, moisture absorption, and fatigue, the initially calibrated viscosity parameters need to be corrected online. The characteristics of the interference signal obtained in real time are used as observations, and the viscosity estimate is dynamically updated using a Kalman filter algorithm.

[0097] In practice, dynamic viscosity is treated as a state variable, with the previous estimate used as a priori prediction. When new detection data is acquired, the observation residual (the difference between the measured signal and the signal predicted based on the current viscosity) is calculated. The Kalman gain is then used to balance the reliability of the prior estimate and the observation information, generating a posterior correction value. This process is repeated in each detection cycle, enabling the viscosity parameter to adaptively track changes in the actual material state and ensuring that damage evolution prediction is always based on the current, true material properties.

[0098] Based on the embodiments provided in this application, initial values ​​are first obtained through the V50 ballistic limit test, a standard test directly related to the ultimate performance of ballistic materials, ensuring that the model parameters have a clear engineering physics background and calibration benchmark. Subsequently, online corrections are performed based on real-time detection data using a Kalman filter algorithm, forming a parameter learning closed loop. This makes the model not merely a static prediction tool, but one capable of continuously updating and optimizing its core parameters using online detection data. This allows it to adapt to differences in materials from different batches or performance degradation during service, maintaining predictive accuracy and enhancing the long-term reliability of the entire intelligent detection system.

[0099] Furthermore, S4, the excitation phase difference and ultrasonic intensity of the shear wave are adjusted in real time according to the spatial gradient of the predicted spatial distribution, including: Calculate the spatial gradient magnitude of the predicted spatial distribution of the interlayer phase field distribution at the next time step; Based on the spatial gradient magnitude, a model prediction and control framework is constructed, with the state space consisting of the interlayer phase field, entropy yield, and orientation gradient, and the action space consisting of the ultrasonic incident angle, intensity, and phase difference. The model predictive control framework is designed to achieve adaptive optimization of the detection process. The state space is set in three dimensions, including the interlayer phase field distribution, interlayer entropy yield, and geometrically necessary orientation gradient. These three state variables collectively describe the damage state, energy dissipation characteristics, and material structure features of the current detection region. The action space includes the ultrasonic incident angle, emission intensity, and phase difference between the two shear waves; these parameters determine the characteristics of the excitation field.

[0100] The predictive model embeds the previously established damage evolution equation, which can predict the state evolution several time steps later based on the current state and control actions. The controller solves a finite-time-domain optimization problem at the current moment to find the optimal sequence of actions that makes the future state follow the expected trajectory (such as the minimum entropy production path), and implements only the first action. Then, it re-optimizes at the next sampling time to form a rolling time-domain control strategy.

[0101] The components of the state space are normalized, including dividing the interlayer phase field by its saturation value, dividing the entropy yield by the characteristic entropy yield, and dividing the orientation gradient by the characteristic curvature, to obtain the state space coordinates. , The coordinate index is 1, 2, or 3. To achieve comparability and numerical stability among different physical quantities, the components of the state space are normalized. The interlayer phase field is divided by its theoretical saturation value (corresponding to a fully layered state), the entropy yield is divided by the characteristic entropy yield (determined based on the material's interlayer shear modulus and typical strain rate), and the orientation gradient is divided by the characteristic curvature (calculated from fiber bundle diameter and typical layup angle). By normalizing these eigenvalues, each state component is mapped to a dimensionless interval, facilitating unified processing in optimization problems.

[0102] The Fisher information matrix, defined by the normalized entropy productivity, is used as the Riemann metric tensor. Constructing a curved manifold in the thermodynamic state space; Choosing the Fisher information matrix of entropy production rate as the Riemannian metric tensor of the state space is based on considerations of information geometry. The Fisher information matrix reflects the uncertainty structure of probability distribution parameter estimation. In this application, entropy production rate, as a key indicator of thermodynamic state, has statistical fluctuation characteristics that define the local geometry of the state space. Using the Fisher information matrix as a metric means that in directions with higher uncertainty in entropy production rate estimation, the distance in the state space is stretched; in directions with higher certainty, the distance is compressed. This geometric structure naturally encodes the statistical correlations between different state variables, providing an inherent uncertainty quantification mechanism for control decisions.

[0103] It should be noted that the normalized entropy yield is not a single value, but rather the expected value of statistical observations constructed based on the time-frequency characteristics of the second intrinsic polarization state (slow wave). Specifically, the energy decay process obtained after the synchronous squeezing transformation of the interferometric signal is regarded as a realization of a stochastic process, and the entropy yield D represents the expected power dissipation per unit volume-temperature of this process. Based on this, using the interlayer phase field (normalized to state-space coordinates), entropy yield, and orientation gradient as parameters to be estimated, the partial derivatives of the log-likelihood function of the observations with respect to the state parameters are constructed. Fisher information matrix. This is the expected covariance of the first-order partial derivative of the log-likelihood function, whose matrix elements quantify the sensitivity of the entropy yield observation to changes in state parameters.

[0104] In the interlaminar damage detection scenario of this application, the Fisher information matrix, as a Riemannian metric tensor, has a clear meaning in quantifying thermodynamic uncertainty: when the eigenvalue of the matrix is ​​large in a certain direction, it indicates that the entropy yield is extremely sensitive to changes in the state parameters in that direction, meaning that the state estimation in that direction has low uncertainty, corresponding to short distances on the state manifold (geodesics tend to pass through such regions to improve detection accuracy); conversely, when the eigenvalue is small (such as when approaching the critical slowing state), it indicates that the entropy yield signal fluctuates violently and parameter estimation is difficult, corresponding to long distances on the manifold (geodesics avoid such high uncertainty regions). Thus, the geometric structure of the curved manifold essentially maps the information identifiable topology within the thermodynamic state space.

[0105] It should be noted that applying Fisher's metric to solve the geodesic equations transforms the control path optimization objective of the model predictive control framework from simply minimizing energy dissipation to a synergistic optimization that maximizes information gain and minimizes thermodynamic costs. That is, natural gradient descent updating along the tangential direction determined by Fisher's metric essentially adjusts the ultrasonic excitation parameters along the direction that maximizes the amount of state information obtained per unit entropy cost. This ensures that in the scanning path planning of the piezoelectric fiber composite transducer array, the control strategy with the strongest discriminative power for interlaminar damage states is prioritized, rather than blindly traversing the parameter space.

[0106] Solve the geodesic equation: The minimum entropy production evolution path is obtained, where Evolutionary parameters along the evolutionary path, State space coordinates For evolution parameters The second rate of change; State space coordinates right The rate of change; State space coordinates right rate of change, To and The corresponding coordinate indices are all 1, 2, and 3; For the Riemannian metric tensor Christoffel notation for calculation, To and Corresponding coordinate index; Among them, the Fisher information matrix is ​​composed of the second derivative of the likelihood function of the entropy production rate with respect to the thermodynamic state space coordinates, and the geodesic equation is solved iteratively in discrete time steps through the Levi-Civita connection simplification form of the Riemannian metric. The optimal adjustment amount in the action space is calculated by updating the tangential direction determined by the Riemann metric through natural gradient descent, and the excitation phase difference and ultrasonic intensity of the shear wave are adjusted in real time.

[0107] Under a given Riemannian metric, the geodesic between two points represents the shortest path connecting them. In a control framework, the minimum entropy production evolution path obtained by solving the geodesic equation represents the ideal trajectory where the entropy production increases least as the system evolves from its current state to the target state. This control philosophy follows the principle of minimum entropy generation, ensuring that the detection process itself minimizes disturbance to the material system while effectively revealing the damage state.

[0108] Natural gradient descent updates parameters using the tangent space structure defined by Riemannian metrics. Unlike ordinary gradient descent in Euclidean space, natural gradient descent considers the geometric curvature of the state space, calculating the optimal adjustment in the action space along the geodesic direction. This method avoids blind searching on curved manifolds, ensuring that the adjustment direction of excitation parameters always conforms to the intrinsic geometry of the thermodynamic state space, thus improving the speed and stability of control convergence.

[0109] Based on the embodiments provided in this application, the design of the control system is elevated to the level of thermodynamic state-space geometry. By placing normalized state variables (phase field, entropy yield, orientation gradient) in a Riemannian space quantified by the Fisher information matrix and solving the geodesic equations on this curved manifold, the aim is to find the evolution path with minimum entropy yield. Specifically, this embodiment no longer relies solely on simple gradient thresholds for feedback control, but attempts to understand and track the natural or most probable evolution trajectory of the material's internal state driven by thermodynamic forces. Based on this trajectory, the optimal excitation parameters (phase difference, intensity) calculated using the natural gradient descent method enable subsequent ultrasonic excitation to more effectively probe the material, potentially acquiring richer state information with lower energy input and reducing unnecessary disturbances to the tested material, thus achieving high efficiency and intelligence in the detection process.

[0110] Furthermore, the excitation phase difference and ultrasonic intensity of the shear wave are adjusted in real time, including: When the spatial gradient of the predicted spatial distribution exceeds the preset phase field gradient threshold, the excitation phase difference of the shear wave is adjusted to the preset phase difference value that maximizes the sensitivity of the interference field to interlayer delamination, and the ultrasonic intensity is increased to the preset enhancement factor of the preset reference value. The preset phase field gradient threshold is set to distinguish different modes of damage evolution. When the spatial gradient of the phase field distribution is below this threshold, the damage is in a relatively stable state, with a slow and controllable evolution rate. When the gradient exceeds this threshold, the damage enters the unstable propagation stage, with a clear localization trend, and may rapidly develop into macroscopic stratification. This threshold is usually estimated using parameters such as the material's interlaminar fracture toughness and fiber volume fraction, or determined based on historical test data, corresponding to the critical point where damage transitions from stable propagation to unstable propagation.

[0111] A preset phase field gradient threshold corresponds to the critical condition for the transition of damage from steady-state propagation to unstable propagation. By preparing standard specimens containing pre-existing delamination defects, the spatial gradient of the interlaminar phase field distribution and the delamination propagation rate are simultaneously recorded during scanning detection. When a nonlinear sharp increase in the delamination propagation rate relative to the gradient is observed (i.e., a sudden increase in the derivative of the propagation rate with respect to the gradient), the corresponding gradient value is recorded as the threshold benchmark. Alternatively, the threshold can be estimated using the theoretical relationship between the material's interlaminar fracture toughness GIC and interlaminar shear modulus G13: the threshold is reached when the energy release rate caused by the phase field gradient exceeds GIC.

[0112] The determination of the preset phase difference value needs to maximize the sensitivity of the interference field to interlayer delamination defects. Through theoretical analysis of the energy distribution of the interference field in the interlayer region under different phase differences, combined with numerical simulation of the response of delamination defects of typical depths and sizes to different phase difference excitations, the phase difference configuration that maximizes the defect-scattered signal is identified. During the experimental calibration phase, different phase difference combinations are scanned on standard specimens with known defects to verify and fine-tune the theoretical predictions, ultimately determining the optimal phase difference setting for delamination at different depths. The method for determining this threshold includes: calculating the standing wave energy distribution of the interference field in the interlayer region under different phase difference excitations based on the propagation theory of shear waves in laminates; calculating the interference enhancement factor of fast and slow shear wave modes at a specific depth (such as the interlayer interface) using transmission line models or finite element simulations, and selecting the phase difference that maximizes the amplitude of particle vibration velocity in the interlayer region. During experimental calibration, on standard specimens containing artificial delaminations of known depths (such as pre-embedded polytetrafluoroethylene films), the range of phase differences from 0 to 2π is traversed, and the point of maximum echo signal amplitude is recorded.

[0113] The determination of the preset enhancement factor includes: based on the signal-to-noise ratio improvement requirements and material safety margin. For every doubling of sound intensity (enhancement factor +1), the echo signal amplitude theoretically increases by 6 dB (signal-to-noise ratio improvement). The intensity increase required to achieve a discernible signal-to-noise ratio (e.g., 20 dB) is calculated based on the detection depth and material attenuation coefficient. Simultaneously, the total enhanced intensity is ensured not to exceed 80% of the material damage threshold. For deep detection (>10 mm) or high-attenuation regions, the enhancement factor is set to 2.0 times (i.e., increased to 1.0-2.0 W / cm²); for shallow or low-attenuation regions, it is set to 1.5 times (i.e., increased to 0.75-1.5 W / cm²). In the control strategy of this embodiment, when the phase-field gradient exceeds this threshold, it automatically switches to an enhancement mode of 1.5 to 2.0 times.

[0114] When the spatial gradient of the predicted spatial distribution is lower than or equal to the preset phase field gradient threshold, the current excitation phase difference is maintained and the ultrasonic intensity is maintained or reduced.

[0115] The setting of the ultrasonic intensity benchmark value needs to comprehensively consider both the detection sensitivity requirements and material safety constraints. The benchmark value is set as the minimum safe intensity while ensuring a sufficient signal-to-noise ratio, avoiding ultrasonic fatigue or thermal damage to the material. When a high-risk region is detected where the phase field gradient exceeds a threshold, the intensity is increased to a preset enhancement factor of the benchmark value. This factor is determined based on the required signal-to-noise ratio increase, while ensuring that the enhanced intensity remains below the ultrasonic damage threshold of the material, avoiding false damage or noise amplification caused by over-excitation. This graded control strategy minimizes interference with equipment detection while ensuring detection reliability. Methods for determining this threshold include: using the maximum safe acoustic intensity at which the material will not experience ultrasonic fatigue or thermal damage as the upper limit; determining the maximum intensity at which the ballistic material under test (especially UHMWPE fiber laminate, which has a low melting point) will not experience a temperature rise exceeding 5°C or fiber / matrix interface debonding under continuous ultrasonic excitation through acoustic emission monitoring or microscopic observation; or setting a safety factor below the material damage threshold based on the transducer output acoustic intensity measured by hydrophone.

[0116] Based on the embodiments provided in this application, a direct correlation is established between the predicted phase field gradient and the excitation parameters: when the prediction indicates that the damage may evolve rapidly (high gradient), the system automatically switches to a preset phase difference that is most sensitive to interlayer delamination and enhances the ultrasonic intensity to focus on observing the risk area; conversely, the excitation intensity is maintained or reduced. This strategy of allocating detection resources on demand allows limited detection energy and time to be focused on the areas most likely to have problems, thereby improving overall detection efficiency and avoiding unnecessary strong excitation in safe areas, which conforms to the principles of economy and safety in nondestructive testing.

[0117] Furthermore, real-time adjustments also include: In regions where the spatial gradient exceeds a preset phase field gradient threshold, the scanning path is densified along the normal direction of the spatial gradient, and the moving step size of the piezoelectric fiber composite transducer array is shortened to a preset compression ratio of the original step size.

[0118] When the spatial gradient of the interlayer phase field distribution exceeds a preset phase field gradient threshold, it indicates a significant localization trend of damage in the region. The gradient direction points towards the direction of damage frontal extension, while the gradient normal direction corresponds to the lateral boundary of the damage region. Intensive scanning along the normal direction can accurately capture the lateral extent of the damage region (i.e., the width of the damage region perpendicular to the gradient direction), avoiding missed detections.

[0119] The determination of the initial step length needs to take into account the balance between detection efficiency and initial resolution. For a typical bulletproof insert (300mm×250mm), the initial step length is usually set to 5mm to 10mm, which corresponds to 1.5 to 2 times the effective beam width of the transducer array, to ensure that there are no blind spots in the initial scan.

[0120] The preset compression ratio is dynamically set based on how much the spatial gradient magnitude exceeds the preset phase field gradient threshold. When the gradient value exceeds the threshold but is less than twice the threshold, the compression ratio is set to 50% (i.e., the step size is shortened to half of the original step size, such as from 10mm to 5mm); when the gradient value exceeds twice the threshold, the compression ratio is set to 25% (i.e., the step size is shortened to 2.5mm). For regions with extremely high gradients (exceeding three times the threshold), the compression ratio can reach 10% (step size 1mm), achieving fine characterization of the damage front.

[0121] Based on the embodiments provided in this application, when a high-risk region (high phase field gradient) is identified, not only are the excitation waveform parameters adjusted, but the transducer array is also automatically controlled to shorten its movement step in the normal direction of that region, resulting in a denser scan. This is equivalent to improving the observation resolution of the region while focusing on key areas. This closed-loop control, which links the spatial sampling rate with the excitation parameters, ensures that finer distribution details can be captured for regions with rapidly changing spatial characteristics, such as the damage evolution front, thereby providing more reliable data for accurately assessing damage morphology and size.

[0122] Furthermore, solving the geodesic equations includes: In the optimization objective of the model predictive control framework, the cost of acoustic entropy production caused by the detection action is introduced as a penalty term. When the entropy yield of the second intrinsic polarization state exhibits a critical slowdown characteristic, the model predictive control is triggered to achieve a high exploration rate, automatically adjusting the shear wave phase difference and reducing the ultrasonic intensity to avoid over-driving the weakly bonded interlayer region.

[0123] The second intrinsic polarization state refers to the second intrinsic polarization state corresponding to the slow shear wave mode, representing the polarization component most sensitive to interlaminar shear stiffness degradation. This characterization quantity runs through polarization whitening separation, entropy yield extraction, and critical slowing determination, forming a complete technical chain from signal detection to adaptive control.

[0124] The acoustic entropy production cost refers to the additional irreversible dissipation induced in the material by the ultrasonic excitation itself. Its quantification is achieved by estimating the ultrasonic energy deposition rate: based on the emitted acoustic intensity of the piezoelectric fiber composite transducer array, the ultrasonic absorption coefficient of the ballistic material under test (predicted using the standard hydrophone method), and the excitation duration, the volumetric heat source power generated during the detection process is calculated. This power is divided by the product of absolute temperature and effective volume to obtain the acoustic entropy production rate increment. In the optimization objective function of the model predictive control framework, the square of this acoustic entropy production rate increment is used as the weight of the penalty term to ensure that the control strategy preferentially selects the detection parameters that minimize the disturbance to the thermodynamic state of the material.

[0125] Critical slowing refers to the phenomenon where the relaxation time of entropy production rate fluctuations returning to equilibrium increases abnormally when the system approaches the critical point of damage and instability. The specific criterion is as follows: calculate the autocorrelation function of the entropy production rate time series of the second intrinsic polarization state within a sliding window (e.g., containing 10 sampling points), and extract the time scale τ required for it to decay to 1 / e. When this time scale τ exceeds twice the historical average, and is simultaneously accompanied by an increase in entropy production rate variance of more than 50% compared to the steady-state value, critical slowing is considered to have occurred.

[0126] When critical slowing is detected, the model predictive control framework switches from greedy optimization mode to high exploration rate mode. Specifically, the search range of the shear wave phase difference is expanded from the neighborhood of the current optimal value (e.g., ±π / 8) to the full range (0 to 2π), and random perturbations (perturbation amplitude ±π / 4) are introduced; at the same time, the ultrasound intensity is forcibly reduced to below 50% of the preset baseline value (e.g., from 1.0 W / cm² to 0.3-0.5 W / cm²), replacing high-power focusing with exploratory low-power scanning to avoid irreversible damage to the interlaminar weak bonding zone in the critical state.

[0127] Based on the embodiments provided in this application, by incorporating the cost of acoustic entropy production as a penalty into the optimization objective, the system proactively weighs information acquisition against potential additional ultrasonic effects (micro-heating or micro-damage) on the material when seeking the optimal detection action, reflecting the pursuit of non-invasiveness in the detection process. More importantly, a warning and action switching mechanism based on the critical slowdown characteristics of entropy production rate is established. When the signal exhibits this characteristic, it indicates that the interlayer bonding may be in an extremely fragile (subcritical) state, thereby triggering a high exploration rate mode and reducing the excitation intensity. By utilizing the characteristics of the detection signal itself as feedback, excessive mechanical driving of the impending failure region is automatically avoided, preventing the detection process itself from accelerating damage, demonstrating the self-protection of the intelligent detection system and its awareness of protecting the tested object.

[0128] Furthermore, S1 also includes: The reference frequency of the shear wave is calculated based on the interlaminar shear modulus, material density, and required detection resolution of the bulletproof material to be tested. When the interlaminar shear modulus is lower than the preset modulus threshold, the preset low-frequency band is selected; when the interlaminar shear modulus is higher than or equal to the preset modulus threshold, the preset high-frequency band is selected.

[0129] A preset modulus threshold is used to distinguish between soft and hard bulletproof materials. It is determined based on the statistical distribution of interlaminar shear modulus in common bulletproof composite materials. When the interlaminar shear modulus is below this threshold, the material exhibits high attenuation and strong dispersion characteristics, requiring low frequencies to ensure penetration depth; conversely, high frequencies are required to ensure resolution.

[0130] The preset modulus threshold is typically set between 1.0 GPa and 1.5 GPa. For soft materials such as fiber laminates (modulus approximately 0.5-0.8 GPa) and aramid fiber laminates (modulus approximately 0.8-1.2 GPa), a preset low-frequency band of 2-5 MHz is selected; for hard materials such as carbon fiber laminates (modulus approximately 3-5 GPa) and ceramic hybrid composites (modulus can exceed 10 GPa), a preset high-frequency band of 5-10 MHz (for shallow high-resolution detection) or a mid-frequency band of 3-6 MHz (while also considering deep penetration) is selected.

[0131] Based on the embodiments provided in this application, a reference frequency is calculated according to the basic physical properties of the ballistic material under test, such as the interlaminar shear modulus and density, and the low-frequency / high-frequency band is switched according to the modulus threshold. The principle is that for softer (low-modulus) materials, using a lower-frequency shear wave can ensure sufficient penetration depth and signal-to-noise ratio; for harder (high-modulus) materials, a higher frequency can be selected to obtain better spatial resolution. This ensures the universality of the method for different types and processes of ballistic materials and lays a good signal foundation for the entire testing process from the outset.

[0132] Furthermore, the method is used for pre-damage detection of in-service bulletproof helmets or bulletproof plates. It identifies weakly bonded interlayer regions that have not yet formed macroscopic stratification through deduction. The weakly bonded interlayer regions are determined as regions where the interlayer phase field distribution meets the preset subcritical damage range.

[0133] The pre-defined subcritical damage range corresponds to an intermediate state where interlayer performance degradation has occurred (such as interface debonding and microcrack initiation), but macroscopic stratification visible to the naked eye has not yet formed. This range was determined using the following experimental-simulation control method: Standard samples with different impact energies (below the macroscopic delamination threshold) were prepared and observed using micro-CT or scanning electron microscopy. The damage area ratio (damaged area / total area) at the interlaminar interface was quantified and statistically analyzed. The spatial distribution of the interlaminar phase field order parameter φ was obtained by inversion. A curve corresponding to the damage area ratio and the average value of the phase field order parameter was established. When the damage area ratio is between 5% and 30% (significant but non-penetrating microscopic damage), the corresponding phase field order parameter φ value is the preset subcritical damage range.

[0134] For typical UHMWPE laminates, the preset subcritical damage range is 0.1 < φ < 0.5; for aramid laminates, due to the different fiber / matrix interface bonding characteristics, the range is adjusted to 0.08 < φ < 0.4. When the interlayer phase field distribution in a certain area of ​​in-service equipment is detected to meet this range, it is judged as a "weak bonding" pre-damage state, and key monitoring or preventive maintenance is recommended.

[0135] Based on the embodiments provided in this application, a specific and important application scenario of this method is the pre-damage detection of in-service protective equipment (such as helmets and plates). Its core value lies in its ability, through a series of techniques, to identify weakly bonded areas between layers that have not yet formed macroscopic stratification. The determination is based on the phase field distribution being within a preset subcritical damage range. This directly positions the detection capability of this method at an early stage of potential problems that is difficult to detect using traditional methods (such as visual inspection, impact testing, and conventional ultrasonic C-scan). For ensuring the reliability of in-service equipment, being able to provide early warning of internal bonding degradation before visible damage such as macroscopic stratification or backplate protrusion has significant safety implications and maintenance value, achieving a leap from damage detection to state prediction and early warning.

[0136] like Figure 3 As shown, in one specific embodiment, the intelligent detection method for anisotropy of bulletproof materials provided by the present invention is implemented through an intelligent closed-loop system that includes sensing, analysis, prediction and control links.

[0137] The system begins with signal excitation and acquisition. A piezoelectric fiber composite transducer array, positioned on the surface of the bulletproof material to be tested, synchronously emits two shear waves with the same frequency, orthogonal polarization directions, and programmable phase difference, according to control commands. These two waves propagate and superimpose within the material, forming a specific ultrasonic interference field. The transducer array synchronously acquires the original wave signal of this interference field.

[0138] Subsequently, the system proceeds to the signal processing and damage inversion steps. In this step, the acquired interferometric signals are first subjected to polarization state analysis to construct a polarization coherence tensor, and independent modal signals corresponding to fast and slow shear waves are separated through eigenvalue decomposition. Next, high-resolution time-frequency analysis is performed on the separated modal signals to extract physical quantities that quantify the internal energy dissipation rate, namely the spatial distribution of interlayer entropy yield, from the energy decay characteristics of the signals. Finally, through a reversible neural network model that incorporates fundamental thermodynamic laws as intrinsic constraints, the entropy yield distribution is inverted into an interlayer phase field distribution and its rate of change over time, which can intuitively and continuously characterize the damage state.

[0139] After obtaining the current damage state, the system enters the damage evolution prediction step. This step takes the inverted current interlayer phase field distribution and the pre-acquired three-dimensional fiber orientation field data of the material obtained through independent measurements as inputs. The spatial variation gradient of the fiber orientation field (i.e., the geometrically necessary orientation gradient) is calculated through mathematical operations, and this gradient parameter is used as a key variable and input into a physically driven damage evolution model based on the phase field method. This model considers the competition mechanism between microscopic elastic energy storage and macroscopic crack propagation energy caused by the fiber orientation gradient, thereby calculating the most likely predictive spatial distribution of the damage phase field at the next moment.

[0140] Based on the prediction results, the system executes the final adaptive control step. This step first calculates the spatial gradient of the predicted phase field distribution. Based on this gradient information, a model predictive controller is activated. This controller performs optimization calculations in an abstract geometric space defined by thermodynamic state variables to solve for the most efficient path for the detection process. According to the solution, the controller generates optimal adjustment commands in real time, dynamically adjusting the excitation wave phase difference, ultrasonic emission intensity, and probe scanning path of the transducer array in the next detection cycle. The adjusted parameters are immediately fed back to the initial signal excitation and acquisition steps, thereby enabling the detection energy and focus of interest to be automatically focused on the most damage-sensitive or evolving regions.

[0141] According to another aspect of the embodiments of this application, an intelligent detection system for the anisotropy of bulletproof materials is also provided. For example... Figure 4 As shown, the system includes: The acquisition unit 41 is used to arrange a piezoelectric fiber composite material transducer array on the surface of the bulletproof material to be tested, and to synchronously emit two shear waves with the same frequency, orthogonal polarization direction and programmable phase difference into the bulletproof material to be tested, thereby forming an interference field inside the bulletproof material to be tested and acquiring the interference signal of the interference field. The calculation unit 42 is used to perform time-frequency decomposition on the interference signal to extract the spatial distribution of interlayer entropy yield, and to calculate and determine the interlayer phase field distribution based on the thermodynamic conjugate relationship between interlayer entropy yield and interlayer damage degree. The deduction unit 43 is used to deduce the predicted spatial distribution of the interlayer phase field at the next moment based on the interlayer phase field distribution and the fiber orientation gradient of the ballistic material to be tested, using a damage evolution model that includes a geometrically necessary orientation gradient dissipation term. The adjustment unit 44 is used to adjust the excitation phase difference and ultrasonic intensity of the shear wave in real time according to the spatial gradient of the predicted spatial distribution.

[0142] It should be noted that the embodiments implemented on the side of the intelligent detection system for anisotropy of bulletproof materials in this application can be referenced to each other, and will not be described in detail in this application.

[0143] According to another aspect of the embodiments of this application, an electronic device for implementing the above-described intelligent detection method for anisotropy of bulletproof materials is also provided. This electronic device may be... Figure 5 The terminal device or server shown. This embodiment uses this electronic device as an example of a server. Figure 5 As shown, the electronic device includes a memory 402, a processor 404, and a transmission device 406. The memory 402 stores a computer program, and the processor 404 is configured to execute the steps of any of the above method embodiments through the computer program.

[0144] Optionally, in this embodiment, the aforementioned electronic device may be located in at least one of a plurality of network devices in a computer network.

[0145] Optionally, the transmission device 406 is used to receive or send data via a network. Specific examples of the network described above may include wired and wireless networks. In one example, the transmission device 406 includes a Network Interface Controller (NIC), which can be connected to other network devices and a router via a network cable to communicate with the Internet or a local area network. In another example, the transmission device 406 is a Radio Frequency (RF) module used to communicate with the Internet wirelessly. Furthermore, the electronic device also includes a display 408 and a connection bus 410, which connects the various module components within the electronic device.

[0146] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for intelligent detection of anisotropy in bulletproof materials, characterized in that, include: S1, arrange a piezoelectric fiber composite material transducer array on the surface of the bulletproof material to be tested, and synchronously emit two shear waves with the same frequency, orthogonal polarization direction and programmable phase difference into the bulletproof material to be tested, forming an interference field inside the bulletproof material to be tested, and collect the interference signal of the interference field; S2, perform time-frequency decomposition on the interference signal to extract the spatial distribution of interlayer entropy yield, and calculate and determine the interlayer phase field distribution based on the thermodynamic conjugate relationship between the interlayer entropy yield and the interlayer damage degree. S3. Based on the interlayer phase field distribution and the fiber orientation gradient of the ballistic material to be tested, the predicted spatial distribution of the interlayer phase field distribution at the next moment is deduced using a damage evolution model that includes a geometrically necessary orientation gradient dissipation term. S4, adjust the excitation phase difference and ultrasonic intensity of the shear wave in real time according to the spatial gradient of the predicted spatial distribution, and repeat S1 to S3 based on the adjusted excitation phase difference and ultrasonic intensity.

2. The intelligent detection method for anisotropy of bulletproof materials according to claim 1, characterized in that, S2, perform time-frequency decomposition on the interference signal to extract the spatial distribution of interlayer entropy yield, and calculate and determine the interlayer phase field distribution based on the thermodynamic conjugate relationship between the interlayer entropy yield and the interlayer damage degree, including: Construct the time-varying Stokes parameters of the interference signal and build the polarization coherence tensor; The polarization coherence tensor is decomposed intrinsically to obtain the first intrinsic polarization state corresponding to the fast shear wave mode and the second intrinsic polarization state corresponding to the slow shear wave mode. The interference signal is then projected onto the first intrinsic polarization state and the second intrinsic polarization state to achieve polarization whitening separation. Synchronous squeezing transformation is performed on the separated first intrinsic polarization state and second intrinsic polarization state respectively. The spatial distribution of the interlayer entropy yield is determined according to the instantaneous energy decay characteristics of each polarization state, and the redistribution bandwidth parameter is adjusted according to the interlayer entropy yield. Based on the thermodynamic conjugate relationship between the interlayer entropy yield and the interlayer damage degree, an initial estimate of the interlayer phase field distribution is calculated and determined.

3. The intelligent detection method for anisotropy of bulletproof materials according to claim 2, characterized in that, The determination of the spatial distribution of interlayer entropy yield includes: The energy decay rate is calculated based on the energy decay characteristics of the second intrinsic polarization state within the time window; Calculate the cumulative energy of the second intrinsic polarization state within the time window. Regarding time The energy decay rate is obtained by differentiation. ; The energy decay rate Divide by the absolute temperature within the time window With effective volume The interlayer entropy productivity is obtained by multiplying the product of the two factors.

4. The intelligent detection method for anisotropy of bulletproof materials according to claim 2, characterized in that, The interlayer phase field distribution is calculated and determined in S2, including: Construct a fiber bundle network diagram, wherein the node positions of the fiber bundle network diagram correspond to the geometric center of each fiber bundle; the weight of the edge connecting adjacent nodes in the fiber bundle network diagram is the product of the interlaminar shear stiffness between adjacent nodes and the square of the cosine of the angle between the fiber orientations of adjacent nodes. Based on the fiber bundle network diagram, the frequency domain Green's function of the simplified Christoffel equation, which retains only the dominant term of interlaminar shear stiffness, is solved to obtain the rough wave propagation kernel.

5. The intelligent detection method for anisotropy of bulletproof materials according to claim 4, characterized in that, The method further includes: A Fourier neural operator is constructed, and the Fourier transform of the rough wave propagation kernel is used as the initial convolution kernel of the neural operator. The learning objective of the neural operator is the residual mapping between the real interference signal and the rough wave propagation prediction.

6. The intelligent detection method for anisotropy of bulletproof materials according to claim 5, characterized in that, The method further includes: A reversible neural network layer is constructed, the reversible neural network layer including an affine coupling structure, and the interlayer phase field order parameter and the interlayer entropy yield are constructed as a bijective mapping pair; In the multiplication branch of the affine coupling structure, a preset multiplication operation node is set to represent the generalized force per unit volume. Multiply by the evolution rate of the interlayer phase field sequence parameter and divide by the absolute temperature The result of the operation is constrained to be equal to the interlayer entropy production rate, and the thermodynamic conjugate relationship is realized through the hard constraint encoding of the preset multiplication operation node. The interlayer phase field distribution and its temporal rate of change are determined by synchronous inversion using hard-constrained coding.

7. The intelligent detection method for anisotropy of bulletproof materials according to claim 1, characterized in that, S3, based on the interlayer phase field distribution and the fiber orientation gradient of the ballistic material to be tested, using a damage evolution model including a geometrically necessary orientation gradient dissipation term, the predicted spatial distribution of the interlayer phase field distribution at the next time step is deduced, including: The three-dimensional fiber orientation field of the ballistic material under test is obtained by polarization-sensitive optical coherence tomography or by shear wave birefringence inversion. ; Calculate the Nye curvature tensor of the three-dimensional fiber orientation field. The geometrically necessary orientation gradient magnitude is obtained. , For the gradient operator, the Nye curvature tensor This constitutes a geometrically necessary orientation gradient; According to the geometrically necessary orientation gradient magnitude Calculate discrete ripple energy storage density and according to The modified interlayer critical strain energy release rate was obtained The discrete ripple energy storage density With interlaminar critical strain energy release rate This constitutes an energy competition term; The evolution is calculated based on the evolution rate equation of the interlayer phase field sequence parameter. Based on the evolution rate and the current time step, the predicted spatial distribution of the interlayer phase field distribution at the next time step is calculated.

8. The intelligent detection method for anisotropy of bulletproof materials according to claim 1, characterized in that, The real-time adjustment of the excitation phase difference and ultrasonic intensity of the shear wave includes: When the spatial gradient of the predicted spatial distribution exceeds the preset phase field gradient threshold, the excitation phase difference of the shear wave is adjusted to a preset phase difference value that maximizes the sensitivity of the interference field to interlayer delamination, and the ultrasonic intensity is increased to a preset enhancement factor of a preset reference value. When the spatial gradient of the predicted spatial distribution is lower than or equal to the preset phase field gradient threshold, the current excitation phase difference is maintained and the ultrasonic intensity is maintained or reduced.

9. The intelligent detection method for anisotropy of bulletproof materials according to claim 8, characterized in that, The real-time adjustment also includes: In the region where the spatial gradient exceeds the preset phase field gradient threshold, the scanning path is densified along the normal direction of the spatial gradient, and the moving step size of the piezoelectric fiber composite transducer array is shortened to a preset compression ratio of the original step size.

10. An intelligent detection system for anisotropy of bulletproof materials, wherein the system implements the intelligent detection method for anisotropy of bulletproof materials as described in claim 1, characterized in that, include: The acquisition unit is used to arrange a piezoelectric fiber composite material transducer array on the surface of the bulletproof material to be tested, and to synchronously emit two shear waves with the same frequency, orthogonal polarization direction and programmable phase difference into the bulletproof material to be tested, thereby forming an interference field inside the bulletproof material to be tested, and to acquire the interference signal of the interference field. The calculation unit is used to perform time-frequency decomposition on the interference signal to extract the spatial distribution of interlayer entropy yield, and to calculate and determine the interlayer phase field distribution based on the thermodynamic conjugate relationship between the interlayer entropy yield and the interlayer damage degree. The deduction unit is used to deduce the predicted spatial distribution of the interlayer phase field at the next moment based on the interlayer phase field distribution and the fiber orientation gradient of the ballistic material to be tested, using a damage evolution model that includes a geometrically necessary orientation gradient dissipation term. The adjustment unit is used to adjust the excitation phase difference and ultrasonic intensity of the shear wave in real time according to the spatial gradient of the predicted spatial distribution.

Images