A laser-ultrasonic inspection system and method for adaptive selection of Lamb wave modes excitation

Abstract

This invention discloses a laser ultrasonic testing system and method for adaptive selection of Lamb wave modes, belonging to the field of laser ultrasonic nondestructive testing. The system includes: an excitation light control link, which uses a spatial light modulator to generate a laser linear array excitation source with dynamically adjustable spatial period, linewidth, and line count; an ultrasonic signal acquisition link, which captures Lamb wave signals using a laser Doppler vibrometer; and a central control host computer, integrating a physical information-driven analysis and control module. Using the Rayleigh-Lamb dispersion equation as a physical constraint, it performs mode decoupling and real-time film thickness inversion on multimode aliasing signals, automatically selects the optimal mode and reconstructs the phase hologram, and drives the spatial light modulator to refresh to achieve adaptive matching between the excitation array and the target mode wavelength. This invention achieves closed-loop adaptive selection and directional enhancement excitation of Lamb wave modes without prior sample knowledge, significantly improving the accuracy and adaptability of nondestructive testing of complex thin film structures.

Description

Technical Field

[0001] This invention belongs to the field of laser ultrasonic nondestructive testing technology, specifically relating to a laser ultrasonic testing system and method with adaptive selection of Lamb wave mode excitation. Background Technology

[0002] Laser ultrasonic testing technology, as a non-contact, broadband, and high spatial resolution non-destructive testing method, plays an increasingly important role in the characterization of high-performance thin films and coating structures in aerospace, advanced manufacturing, and defense industries. The core principle of this technology lies in using high-energy pulsed lasers to induce the thermoelastic effect on the material surface, generating ultrasonic signals containing rich information about the material's internal structure. By analyzing the signal's propagation behavior in the medium, physical characteristics such as material thickness, elastic modulus, and internal defects can be characterized. Among these, Lamb waves, as guided waves propagating in thin plates and layered structures, have become a key acoustic carrier for micron- and even submicron-level coating inspection due to their concentrated energy within the waveguide structure and extremely high sensitivity to interfacial delamination, microcracks, and geometric parameter fluctuations.

[0003] However, the inherent multimodal coupling and nonlinear dispersion characteristics of Lamb waves pose challenges to signal aliasing and feature extraction in detection applications. When the wavelength of the excitation source satisfies the spatial coherence condition with the wavelength of a specific Lamb wave mode, selective enhancement excitation of that mode can be achieved, thereby significantly improving the signal-to-noise ratio and modal purity. Existing technologies for achieving mode-selective excitation of Lamb waves mainly fall into the following categories: The first category is the physical mask or diffraction grating scheme, which converts the excitation laser into a linear array with a specific period by setting a mask or grating with a fixed spatial period in the optical path. The characteristic parameters of the first type of scheme are physically locked during processing and cannot be adjusted online. When facing samples of unknown or variable thickness, a dynamic mismatch will inevitably occur between the excitation wavelength and the actual physical properties of the test area. The second type is the scheme that uses a spatial light modulator (SLM) to generate a periodic excitation source. For example, the device proposed by Grünsteidl et al., which generates an adjustable spatial source distribution based on SLM-shaped excitation laser spot, has achieved digital control of array parameters. However, the parameter setting still depends on the preset values ​​of known material thickness and elastic constants. Essentially, it is still an open-loop control and does not have the ability to make autonomous decisions based on signal feedback. The third type is the moving continuous laser source scheme, which achieves selective excitation by matching the laser source moving speed with the phase velocity of the target mode through mechanical scanning. This method relies on mechanical motion and has limitations in terms of response speed, control flexibility and system integration. Furthermore, in terms of signal processing and physical parameter inversion, existing technologies widely employ iterative optimization methods such as genetic algorithms and simulated annealing, which are computationally time-consuming and difficult to meet the requirements of online real-time detection. Although Physical Information Neural Network (PINN) has been initially applied to laser ultrasonic field modeling, it mainly focuses on the forward modeling and inverse parameter derivation of single-mode surface waves, and has not yet involved the decoupling of multi-mode Lamb wave aliasing signals and the real-time joint inversion of film thickness parameters.

[0004] In summary, existing technologies suffer from the following fundamental technical shortcomings when dealing with online detection of thin films / coatings of unknown thickness, varying thickness, and heterogeneous thickness: First, the setting of excitation parameters heavily relies on prior knowledge, making it impossible to autonomously determine the optimal excitation conditions under "double unknown" conditions where both thickness and material constants are lacking. Second, the signal decoupling and parameter inversion stages are isolated from the excitation control stages, failing to form a closed-loop "perception-decision-execution" chain. Third, the lack of real-time perception of local physical property changes and adaptive correction mechanisms for excitation parameters during full-field scanning limits the consistency and spatial resolution of detection results. These shortcomings constitute a pressing technical challenge in this field, necessitating a laser ultrasonic detection scheme capable of adaptively controlling excitation parameters under "double unknown" conditions. Summary of the Invention

[0005] This invention aims to address the following technical problems of existing laser-ultrasonic Lamb wave detection technology when dealing with thin films / coatings of unknown thickness, varying thickness, and heterogeneous thickness: the excitation parameter setting heavily relies on prior knowledge, making it impossible to autonomously determine the optimal excitation conditions under "double unknown" conditions where thickness and material constant information are lacking; the signal decoupling, parameter inversion, and excitation control stages are isolated from each other, failing to form a closed-loop "perception-decision-execution" chain; and the lack of real-time perception of local physical property changes and adaptive correction mechanism for excitation parameters during full-field scanning results limits the consistency and spatial resolution of the detection results.

[0006] The purpose of this invention is to provide a laser ultrasonic testing system and method with adaptive selective excitation of Lamb wave modes. By combining spatial light phase modulation technology, physical information-driven intelligent analysis and closed-loop adaptive control, it can achieve adaptive, high-precision, full-field non-destructive testing of thin films and coatings with unknown thickness and variable thickness.

[0007] To achieve the above-mentioned objectives, the present invention provides the following technical solution:

[0008] In a first aspect, the present invention provides a laser ultrasonic detection system with Lamb wave mode adaptive selection excitation, comprising: an excitation light modulation link, an ultrasonic signal acquisition link, a central control host computer, and a unified time reference link.

[0009] The excitation beam control chain includes a pulsed laser, a beam homogenization component, a spatial light modulator (SLM), and a 4f lens group. The SLM, controlled by a central control computer, reconstructs the continuous incident laser into a laser linear array excitation source with dynamically adjustable spatial period, linewidth, and line count by pixel-scale phase modulation of the incident laser. This source acts on the surface of the sample to excite a Lamb wave signal. The beam homogenization component includes a beam expander and a diffractive optical homogenizer (DOE). The DOE, positioned before the SLM, shapes the incident pulsed laser beam into a flat-topped spot with uniform spatial intensity distribution, eliminating clutter interference introduced by the Gaussian energy gradient and ensuring energy uniformity of the excitation array from its physical source. The 4f lens group, positioned after the SLM, performs spectral spatial filtering and projection scaling on the modulated excitation beam, filtering out higher-order diffraction stray orders and ensuring high contrast and wavefront purity of the excitation linear array.

[0010] The ultrasonic signal acquisition link includes a laser Doppler vibrometer and a multi-degree-of-freedom high-precision displacement stage. The laser Doppler vibrometer is used to non-contactly capture the Lamb wave time-domain signals of each detection point on the surface of the sample and transmit them to the central control host computer. The multi-degree-of-freedom high-precision displacement stage is used to drive the switching of detection points and synchronously feed back position coordinate information. It is preferably a two-degree-of-freedom high-precision displacement stage with a repeatability accuracy better than 0.5 μm to eliminate the interference of mechanical drift on signal inversion.

[0011] The central control host computer integrates a physical information-driven analysis and control module, which is the core of adaptive closed-loop control. Its working mechanism is as follows: First, based on the physical constraints established by the Rayleigh-Lamb dispersion equation, modal decoupling and feature extraction are performed on the Lamb wave time-domain signal acquired by the ultrasonic signal acquisition link, at least retrieving the local film thickness parameters of the current detection point. Specifically, the central control host computer is pre-loaded with a physical information neural network (PINN) trained with physical constraints. This network uses the Rayleigh-Lamb dispersion equation as a regularized physical constraint term in the loss function to decouple the multimodal aliased Lamb wave signal contained in the acquired single-point original time-domain signal into independent modal components, and outputs the film thickness parameters and modal decoupling information of the current detection point in real time. Second, based on the modal decoupling information, the modal purity index and signal-to-noise ratio index of each independent modal component are quantitatively extracted. The optimal detection mode of the current detection point is automatically selected from the multimodal coexistence field through a weighted evaluation function, and the target wavenumber corresponding to this optimal detection mode is extracted. Subsequently, based on the target wavenumber of the selected optimal detection mode and the inverted local film thickness parameters, the phase hologram of the spatial light modulator is reconstructed in real time, and the spatial period and linewidth of the laser line array excitation source are dynamically adjusted to match the wavelength of the target detection mode, thereby achieving coherent enhancement excitation of the mode at the physical level.

[0012] The central control host computer, spatial light modulator, and laser Doppler vibrometer form an adaptive closed-loop control link. This closed-loop control link is equipped with online iterative optimization logic: during the full-field scanning detection process of the multi-degree-of-freedom high-precision displacement stage, when the signal characteristic parameters of the current detection point—including at least one of modal purity, signal-to-noise ratio, and inversion residual—are found to be below a preset threshold, the central control host computer automatically activates the parameter optimization program and re-executes the film thickness inversion, optimal mode locking, and phase hologram refresh based on the ultrasonic signal of the current detection point until the signal characteristic parameters of the current detection point recover to above the preset threshold, thereby ensuring that all detection points in the entire field are in the optimal excitation state.

[0013] The unified time reference link is used to provide a unified time reference for the laser excitation time of the excitation light control link and the signal acquisition time of the ultrasonic signal acquisition link, ensuring the phase resolution accuracy of the multimodal dispersion characteristics. It is a physical prerequisite for performing Lamb wave dispersion analysis and wave field image reconstruction.

[0014] Preferably, the beam homogenization component includes a beam expander and a diffractive optical homogenizer. The diffractive optical homogenizer is disposed in the optical path before the spatial light modulator and is used to shape the incident pulsed laser beam into a flat-topped light spot with uniform intensity distribution in space, so as to eliminate clutter interference introduced by the Gaussian energy gradient.

[0015] Preferably, the central control host computer is preloaded with a physical information neural network (PINN) trained with physical constraints. The PINN uses the Rayleigh-Lamb dispersion equation as the regularized physical constraint term of the loss function to decouple the multimodal aliased Lamb wave signal contained in the original time domain signal of the collected single point into independent modal components, and outputs the film thickness parameters and modal decoupling information of the current detection point in real time.

[0016] Preferably, the physical information-driven analysis and control module in the central control host computer is further configured to: based on the modal decoupling information output by the PINN, quantitatively extract the modal purity index and signal-to-noise ratio index of each independent modal component, automatically select the optimal detection mode of the current detection point from the multimodal coexistence field through a weighted evaluation function, and extract the target wavenumber corresponding to the optimal detection mode.

[0017] Preferably, the physical information driven analysis and control module in the central control host computer reconstructs the phase hologram of the spatial light modulator in real time based on the target wavenumber of the selected optimal detection mode and the inverted local film thickness parameters, and dynamically adjusts the spatial period and linewidth of the laser line array excitation source so that the spatial period matches the wavelength of the target detection mode, thereby realizing the physical coherent enhancement excitation of the mode.

[0018] Preferably, the central control host computer, the spatial light modulator, and the laser Doppler vibrometer form an adaptive closed-loop control link, and the adaptive closed-loop control link is configured with online iterative optimization logic.

[0019] During the full-field scanning detection process of the multi-degree-of-freedom high-precision displacement stage, when the signal characteristic parameters of the current detection point are detected to be lower than the preset threshold, the central control host computer automatically activates the parameter optimization program and re-executes the film thickness inversion, optimal mode locking and phase hologram refresh based on the ultrasonic signal of the current detection point until the signal characteristic parameters of the current detection point are restored to above the preset threshold.

[0020] Preferably, the signal characteristic parameters include at least one of modal purity, signal-to-noise ratio, and inversion residual.

[0021] Preferably, the multi-degree-of-freedom high-precision displacement stage is a two-degree-of-freedom high-precision displacement stage with a repeatability accuracy better than 0.5μm.

[0022] Secondly, this invention provides a laser ultrasonic detection method with adaptive selection of Lamb wave modes for excitation, implemented based on the above-mentioned system, comprising the following steps:

[0023] S1: Drive the multi-degree-of-freedom high-precision displacement stage to the current detection coordinates. The central control host computer drives the spatial light modulator to generate an initial linear array hologram, forming a laser linear array excitation source on the surface of the sample to be tested. The original Lamb wave time domain signal of the current detection point is obtained through a laser Doppler vibrometer.

[0024] S2: Utilizing the pre-loaded physical information-driven analysis and control module in the central control host computer, based on the physical constraints established by the Rayleigh-Lamb dispersion equation, modal decoupling is performed on the original Lamb wave time-domain signal. The local film thickness parameters at the current detection point are inverted, and the optimal detection mode for the current detection point is selected from the decoupled multimodal components by combining signal characteristic parameters. Specifically, this step includes: mapping the acquired original Lamb wave time-domain signal to a two-dimensional time-frequency plane via time-frequency transformation; calling the pre-trained physical information neural network, using the Rayleigh-Lamb dispersion equation as the physical constraint regularization term of the loss function, searching for the globally optimal solution that minimizes the residual in the solution space, and simultaneously outputting each independent modal component after decoupling and its corresponding local film thickness parameters.

[0025] S3: Extract the target wavenumber corresponding to the optimal detection mode, reconstruct the phase hologram in real time and refresh the spatial light modulator, dynamically adjust the spatial parameters of the laser line array excitation source to match the wavelength of the target detection mode, realize the directional enhancement excitation of the mode, and complete the signal recording of the current detection point.

[0026] S4: Drive the multi-degree-of-freedom high-precision displacement stage to switch to the next detection point. The system uses the optimal excitation parameters of the previous detection point to obtain the signal of the current detection point and executes adaptive branch logic judgment: If the signal characteristic parameters and inversion residual of the current detection point are better than the preset threshold, the detection result is directly recorded and the process jumps to step S5; If the signal characteristic parameters or inversion residual of the current detection point are worse than the preset threshold, the closed-loop optimization program is automatically triggered, and the process returns to step S2 to re-execute film thickness inversion, mode locking and phase hologram refresh.

[0027] S5: Traverse all detection points within the test area, summarize the local film thickness parameters and ultrasonic feature values ​​extracted from each detection point under the optimal excitation state, and reconstruct the global structural distribution of the sample under test.

[0028] Preferably, the modal decoupling and film thickness inversion performed in step S2 by using the analysis and control module driven by the pre-loaded physical information in the central control host computer specifically includes:

[0029] The acquired raw Lamb wave time-domain signal is mapped to a two-dimensional time-frequency plane through time-frequency transformation;

[0030] The pre-trained physical information neural network is invoked, and the Rayleigh-Lamb dispersion equation is used as the physical constraint regularization term of the loss function to search for the global optimal solution that minimizes the residual in the solution space. At the same time, the decoupled independent modal components and the corresponding local film thickness parameters are output.

[0031] Compared with the prior art, the technical solution of the present invention has the following beneficial effects.

[0032] 1. Overcoming the detection barrier under "dual unknown" conditions and eliminating dependence on prior knowledge. This invention utilizes a pre-loaded physical information neural network, with the Rayleigh-Lamb dispersion equation as the physical constraint regularization term, to directly decouple multimodal aliasing signals from the single-point raw time-domain signal and invert local film thickness parameters in real time. This mechanism enables the system to autonomously complete online sensing of physical parameters without any prior knowledge about sample thickness or material constants, fundamentally solving the problem of high dependence on prior sample knowledge in traditional laser ultrasonic Lamb wave detection, and providing a reliable physical benchmark for the precise control of subsequent excitation parameters.

[0033] 2. Achieving adaptive mode selection and dynamic physical matching of the excitation optical field significantly improves signal quality. This invention quantifies and extracts features such as mode purity and signal-to-noise ratio based on mode decoupling results. It automatically selects the optimal detection mode from a multi-modal coexisting field using a weighted evaluation function, and dynamically reconstructs the SLM phase hologram accordingly. This ensures precise alignment of the spatial period of the excitation linear array with the target mode wavelength, achieving physically-level directional enhancement excitation. Compared to existing technologies that blindly select modes based solely on signal strength or rely on fixed parameters for excitation, this invention achieves target mode locking and enhancement at the transmitting end, effectively suppressing stray mode interference and surface scattering noise, and significantly improving the mode purity and signal-to-noise ratio of the signal.

[0034] 3. Constructing an online iterative closed loop for full-field scanning to ensure high consistency of detection results. This invention embeds a closed-loop optimization mechanism based on signal characteristic parameters during the full-field scanning process on the displacement stage. When signal degradation is detected at the current measurement point, the system automatically re-executes the film thickness inversion, modal locking, and hologram refresh processes without manual intervention. This mechanism effectively compensates for the physical property drift caused by non-uniform coating degradation and geometric irregularities, ensuring that all detection points are in the optimal excitation state, guaranteeing high consistency of scan data in terms of signal-to-noise ratio and modal purity, thereby improving the spatial resolution of damage imaging and the accuracy of quantitative analysis.

[0035] 4. Ensuring excitation quality from the optical source and enhancing the physical limits of coherent enhancement. This invention employs a cascaded optical link of "DOE flat-top homogenization + SLM phase modulation + 4f spatial filtering" to eliminate clutter interference introduced by the Gaussian beam energy gradient before excitation and filter out diffraction stray orders, ensuring the generation of a high-contrast, energy-uniform, and wavefront-pure excitation linear array. This optical link provides a precise hardware execution basis for the physically driven adaptive control scheme, allowing the modal coherent enhancement effect to be fully utilized and improving the overall detection performance of the system. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the structure of the Lamb wave mode adaptive selection excitation laser ultrasonic detection system provided in an embodiment of the present invention.

[0037] Figure 2 This is a flowchart illustrating the analysis process for inverting Lamb wave thin layer thickness and modal information in an embodiment of the present invention.

[0038] Figure 3 This is an overall flowchart of the ultrasonic signal scanning and extraction method in an embodiment of the present invention.

[0039] The meanings of the symbols in the attached icons are as follows:

[0040] 1-ns pulsed laser, 2-beam expander, 3-diffractive optical homogenizer, 4-spatial light modulator, 5-4f lens group, 6-unified time reference link, 7-two-degree-of-freedom high-precision displacement stage, 8-laser Doppler vibrometer, 9-central control host computer, 10-excitation array, 11-target Lamb wave acoustic field, 12-scanning array. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0042] It should be noted that the terms "first," "second," etc., used in the specification and claims of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention can be implemented in orders other than those illustrated or described. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion.

[0043] This invention provides a Lamb wave mode-adaptive selective excitation laser ultrasonic detection system. For example... Figure 1 As shown, the system can be functionally divided into four main components: the excitation light control link, the ultrasonic signal acquisition link, the central control host computer, and the unified time reference link. Figure 1 As shown in the physical layout, the system includes: an ns pulsed laser 1, a beam expander 2, a homogenizing diffraction optical element 3, a reflective spatial light modulator 4, a 4F lens group 5, a unified time reference link 6, a two-degree-of-freedom high-precision displacement stage 7, a laser Doppler vibrometer 8, and a central control host computer 9. The pulsed laser output from the ns pulsed laser 1 passes sequentially through the beam expander 2, the homogenizing diffraction optical element 3, the reflective spatial light modulator 4, and the 4F lens group 5, forming a preset excitation array 10 on the surface of the sample under test, thereby exciting a Lamb wave acoustic field 11 in the sample. The laser Doppler vibrometer 8 receives the Lamb wave signal and, driven by the two-degree-of-freedom high-precision displacement stage 7, acquires the ultrasonic signal point-by-point along the preset trajectory of the scanning array 12. The unified time reference link 6 provides a strictly synchronized clock reference for the excitation time of the ns pulsed laser 1 and the acquisition time of the laser Doppler vibrometer 8. The central control host computer 9 serves as the core of the system's control and processing. On the one hand, it receives signals collected by the laser Doppler vibration meter 8 and performs intelligent analysis. On the other hand, it adjusts the phase hologram of the reflective spatial light modulator 4 in real time based on the analysis results, thus forming a complete adaptive closed-loop detection circuit.

[0044] The following section provides a detailed explanation of the specific composition, working principle, and collaborative relationship of the four major functional components mentioned above.

[0045] The excitation light modulation link is the physical basis for realizing tunable excitation and directional enhancement of Lamb wave modes. Starting from the light source, the link consists of an ns pulsed laser 1, a beam expander 2, a homogenizing diffraction optical element 3, a reflective spatial light modulator 4, and a 4f lens group 5. The components are cascaded and coupled through free space optical paths.

[0046] A nanosecond pulsed laser 1 serves as the excitation source for the system, preferably a Q-switched solid-state laser, such as an Nd:YAG laser. The output pulse width of this laser is strictly controlled to the nanosecond level, for example, less than 8 ns. This extremely short pulse duration ensures a high concentration of laser energy along the time axis. On the one hand, this facilitates the excitation of a broadband Lamb wave signal on the material surface, providing a rich frequency selection range for subsequent screening and locking of specific modes in the wavenumber domain. On the other hand, the short pulse width also ensures precise timing alignment between the excitation and acquisition times. The laser's operating wavelength is preferably in a band where the material has high light absorption efficiency, such as 1064 nm or its harmonic wavelengths. The excitation process is based on the thermoelastic effect principle, where the laser energy density is controlled below the damage threshold of the material under test. Laser irradiation causes localized instantaneous thermal expansion on the material surface, thereby non-destructively exciting a broadband Lamb wave signal. This excitation method has good repeatability and requires no coupling agent, fully meeting the requirements of non-contact in-situ detection.

[0047] The original laser beam output from the ns pulsed laser 1 typically exhibits a Gaussian distribution in its cross-section, meaning that the energy density is high in the central region and low in the peripheral region. If this non-uniform energy distribution is directly transmitted to the subsequent linear array generation stage, it will lead to significant energy differences between laser lines at different positions in the generated excitation array 10, thereby introducing clutter interference at the Lamb wave excitation source and reducing the coherence enhancement purity and signal-to-noise ratio of the target mode. Therefore, the excitation light control link of this invention includes a beam homogenization component consisting of a beam expander 2 and a homogenizing diffraction optical element 3 before the beam enters the spatial light modulator.

[0048] Specifically, beam expander 2 first receives the raw pulsed laser output from ns pulsed laser 1 and expands the beam cross-sectional area through its internal lens group, ensuring that the expanded spot size precisely matches the effective pixel area of ​​the subsequent reflective spatial light modulator 4. This beam expansion process not only ensures that the spatial light modulator 4 can receive complete incident wavefront information but also effectively reduces the instantaneous peak power density per unit area on the liquid crystal panel of the spatial light modulator 4 by increasing the light-receiving area of ​​the spot, ensuring that it always operates within a safe range below the damage threshold. The expanded laser beam is then incident on homogenizing diffraction optical element 3. This element utilizes the principle of diffraction optics, with precisely designed microstructure patterns on its surface, which can accurately redirect the spatial distribution of incident photons, shaping the beam with Gaussian distribution characteristics into a flat-top spot with a highly uniform intensity distribution within a specific working cross-section. After homogenization, the energy density difference between the central and edge regions of the flat-top beam is significantly eliminated, thus ensuring that the thermoelastic excitation intensity generated by each laser line on the surface of the sample in the subsequently generated linear array excitation source remains highly consistent. A uniform linear array can ensure that the interference effect of each sub-wave source is optimized in the wavenumber domain, thereby improving the purity and signal-to-noise ratio of the target Lamb wave mode from the source. This is the primary physical guarantee for achieving high-fidelity mode-selective excitation.

[0049] The flat-top beam, homogenized, then illuminates the liquid crystal panel of the reflective spatial light modulator 4. The reflective spatial light modulator 4 is the core actuator for fully digitally controlling the geometric parameters of the excitation source. Its working principle involves loading a phase hologram generated by a specific algorithm onto the central control host computer 9, applying corresponding phase delay compensation at the pixel scale to each point on the incident light wavefront, thereby changing the phase gradient distribution of the reflected light field. Macroscopically, this wavefront phase modulation process reconstructs the originally continuous flat-top beam into a linear array structure with specific spatial geometric features. Compared to traditional physical masks or metal grating schemes, the use of the spatial light modulator 4 enables the present invention to dynamically adjust various geometric parameters of the excitation array in real time without replacing any hardware. Specifically, the central control host computer 9 can independently and decoupledly control three core parameters of the generated linear array by refreshing the phase hologram: spatial period Δs, the distance between the centers of two adjacent laser lines; linewidth w, the width of a single laser line; and the number of lines N, the total number of laser lines in the linear array. The decoupling independence between parameters means that the system can independently optimize the above parameters according to actual detection needs, such as different factors such as the surface roughness of the sample to be tested, the coating thickness distribution range, and the target modal wavelength, so as to achieve highly flexible customized configuration of the excitation source.

[0050] The modulated excitation array beam is output from the reflective spatial light modulator 4 and enters the 4f lens group 5. The 4f lens group 5 consists of two lenses with focal lengths of f1 and f2, forming a classic 4f optical imaging system in the optical path. This lens group performs a dual function in this invention. Firstly, it performs a spatial filtering function on the spectral plane. On the Fourier spectral plane of the 4f system, the zero-order high-brightness spot and higher-order stray diffraction orders generated by the pixelated structure diffraction of the spatial light modulator 4 are spatially separated from the diffraction orders carrying useful excitation array information. This invention utilizes this physical characteristic by setting a physical aperture on the spectral plane, selectively blocking the zero-order light and higher-order stray orders through low-pass or band-pass filtering, allowing only the target diffraction orders to pass. After this filtering process, a clean linear array structure with sharp edges and extremely high contrast can be recovered at the output of the 4f system, eliminating the interference of stray light on the excitation effect. Secondly, it performs a projection scaling function. By selecting or adjusting the focal length ratio f2 / f1 of the two lenses, precise scaling control can be achieved on the projection size of the excitation array pattern onto the surface of the sample under test. This scaling mechanism allows the micron-level physical dimensions of the excitation array to be flexibly adjusted according to actual detection needs, optimizing the energy convergence density of the excitation light on the sample surface, while ensuring that the acoustic pressure amplitude and directivity of the target single-mode Lamb wave excitation meet the detection requirements. The excitation light, after being processed by the 4f lens group 5, is finally focused onto the surface of the sample under test, forming a pattern as shown in the image. Figure 1 The high-quality excitation array shown in Figure 10 generates a Lamb wave acoustic field 11 inside the sample under the thermoelastic effect. Since the spatial period Δs of the excitation array determines the phase relationship between each sub-wave source, when Δs satisfies the spatial coherence enhancement condition with the wavelength of a specific Lamb wave mode, that mode will be directionally enhanced in the wavenumber domain, while other modes will be suppressed due to phase mismatch. This achieves the physical selection and enhancement of the target mode by the excitation source.

[0051] The ultrasonic signal acquisition link is responsible for acquiring the weak physical vibration signals caused by the propagation of Lamb waves on the surface of the sample under test and converting them into digital signals that can be analyzed. This link mainly consists of two parts: a laser Doppler vibrometer 8 and a two-degree-of-freedom high-precision displacement stage 7.

[0052] The laser Doppler vibrometer 8, as the core signal acquisition device of the system, performs non-contact measurement of minute vibration displacements or velocities on the sample surface based on the optical Doppler effect. In this invention, considering the wide spectral coverage of Lamb wave multi-modes in thin-layer materials and the extremely weak vibration amplitudes corresponding to high-frequency modes, the selected laser Doppler vibrometer 8 preferably possesses high bandwidth and high resolution performance indicators. Specifically, its sampling frequency can reach up to the order of 50MHz, which can completely cover the spectral measurement requirements of all Lamb wave modes, ensuring that no amplitude attenuation or phase distortion occurs during signal acquisition; its displacement resolution can reach the sub-nanometer level, which is sufficient to distinguish the acoustic response differences caused by minute changes in coating thickness. The above performance indicators provide low-distortion, high-fidelity raw signals for subsequent time-frequency domain analysis, modal decoupling, and physical parameter inversion. In addition, the selected laser Doppler vibrometer 8 preferably integrates a self-focusing function. When performing full-field multi-point scanning in conjunction with the two-degree-of-freedom high-precision displacement stage 7, it can automatically compensate for defocusing caused by changes in sample surface roughness or curvature, ensuring that the ultrasonic signals collected at each detection coordinate point have high comparability and consistency in terms of phase, spectrum and amplitude information.

[0053] The two-degree-of-freedom high-precision displacement stage 7 carries the probe head of the laser Doppler vibrometer 8, performing precise motion along a preset gridded scanning trajectory within the test area. This displacement stage provides high-precision displacement motion in two mutually orthogonal degrees of freedom—for example, along the x and y directions of the sample surface plane—with a repeatability preferably better than 0.5 μm. During the detection process, the displacement stage 7 moves the probe head stepwise to each preset detection coordinate point in the scanning array 12 according to coordinate commands issued by the central control host computer 9, and completes the acquisition of the Lamb wave signal at each measurement point during its dwell time. The high-precision positioning characteristics of the displacement stage effectively eliminate the interference introduced by mechanical drift in the inversion calculation of key parameters such as ultrasonic time-of-flight (TOF) and phase velocity, forming the mechanical basis for ensuring the spatial consistency and quantitative accuracy of the full-field scanning data.

[0054] The central control host computer 9 is the decision-making core and intelligent hub of the entire detection system. It integrates a physical information-driven analysis and control module, realizing a complete closed-loop control logic from signal perception and intelligent decision-making to light field reconstruction. The workflow of this analysis and control module can be found in [reference needed]. Figure 2 Understand the inversion analysis process shown.

[0055] After the laser Doppler vibrometer 8 completes the acquisition of the original Lamb wave time-domain signal at a certain detection coordinate point and transmits the data back to the central control host computer 9, the analysis and control module first performs a joint time-frequency domain analysis on the signal. Specifically, the module uses time-frequency analysis techniques such as short-time Fourier transform and continuous wavelet transform to map the one-dimensional time-domain signal s(t) to a two-dimensional time-frequency complex plane, thereby extracting the instantaneous amplitude evolution characteristics and phase evolution characteristics of the signal. This time-frequency transformation step allows different modes of Lamb waves to exhibit distinguishable energy distribution trajectories in the time-frequency domain due to their different dispersion characteristics, providing an initial feature space for subsequent mode separation.

[0056] Subsequently, the analysis and control module invokes a pre-trained and deployed physical information neural network model to perform modal decoupling and physical parameter inversion on the time-frequency transformed ultrasonic signal. The core innovation of this physical information neural network lies in the embedding of strict physical constraints during its training and inference processes—specifically, using the Rayleigh-Lamb dispersion equation as a regularization constraint term in the loss function. The Rayleigh-Lamb dispersion equation describes the implicit functional relationship between frequency f, wave number k, and phase velocity when a Lamb wave propagates in a thin plate of given thickness and material constants; mathematically, it can be represented as Ω(f,k,h)=0. By directly introducing this physical law into the neural network's training objective function, the network, during optimization, not only needs to fit the training data but must also strictly adhere to the solution space boundary conditions specified by the physical law. This physical constraint mechanism endows the model with extremely strong generalization ability and physical consistency. In the actual detection and inference phase, for the input single-point raw ultrasonic signal, this physical information neural network can automatically search for the globally optimal solution that minimizes the physical residual within the solution space defined by the Rayleigh-Lamb dispersion equation, thereby simultaneously achieving two key outputs: first, decoupling the multimodal aliased Lamb wave signal contained in the raw signal into mutually independent modal components; second, retrieving in real time the local film thickness parameter h that best matches the current signal dispersion characteristics. In particular, because this physical information neural network operates using a pre-trained model for forward inference, it eliminates the need for complex on-site numerical iteration calculations during each detection process, thus meeting the stringent response speed requirements of online real-time detection.

[0057] After obtaining the modal decoupling results and local film thickness parameters, the analysis and control module further performs modal feature quantification evaluation and automatic optimal mode selection. The module quantifies and extracts multiple performance indicators for each independent modal component from the high-dimensional decoupling features output by the neural network, including at least modal purity and signal-to-noise ratio (SNR). Modal purity characterizes the concentration of the target modal component's energy within the total signal energy, reflecting the cleanliness of modal separation; the SNR characterizes the relative strength of the target modal signal intensity relative to background noise and environmental interference. The analysis and control module uses these indicators to construct a weighted evaluation function, comprehensively ranking each candidate mode in a multimodal coexistence field, and automatically selecting the optimal detection mode with the highest detection gain and best physical fit at the current detection coordinate point. Compared to the empirical approach of traditional laser ultrasound detection that selects modes solely based on signal amplitude intensity, this intelligent selection mechanism based on multidimensional quantification indicators effectively avoids the risk of misselecting interfering modes in high-noise environments, fundamentally enhancing the physical accuracy of Lamb wave analysis.

[0058] After selecting the optimal detection mode, the analysis and control module automatically calculates the optimal linear array spatial parameters required to achieve physical coherent enhancement excitation of that mode, based on the physical wavenumber k corresponding to that mode and the locally obtained film thickness parameter h. This reasoning process is based on the physical laws revealed by the Lamb wave dispersion relation: for a given film thickness h and target mode wavenumber k, there exists a definite optimal excitation spatial period Δs, such that the excitation phases of each sub-wave source satisfy the constructive interference condition in the propagation direction of the target mode. The analysis and control module calculates the optimal spatial period Δs and linewidth w, and generates the corresponding phase hologram through reverse mapping, sending the command to the reflective spatial light modulator 4 for refresh. The spatial light modulator 4 dynamically adjusts the phase grating structure it generates through high-speed phase refresh, thereby ensuring a high degree of matching between the spatial period Δs of the excitation array 10 and the wavelength λ of the target Lamb wave mode. This process constitutes a complete "perception → decision → reconstruction" closed loop: the laser Doppler vibrometer 8 senses the signal → the physical information neural network decouples and analyzes the signal to determine the optimal mode and parameters → the spatial light modulator 4 reconstructs the excitation light field based on the new hologram. Closed-loop control enables the excitation source to adaptively adjust in real time according to the local physical characteristics of each detection point on the sample, achieving locked excitation of the target single-mode Lamb wave at the physical level. This effectively suppresses scattering noise caused by surface roughness and significantly improves the adaptability of laser ultrasonic testing to complex thin-layer structures.

[0059] A key feature of this invention is that the closed-loop control described above does not only function in the initial excitation stage of single-point detection, but rather persists throughout the entire field-wide scanning detection process, forming a continuously operating online iterative optimization mechanism. The specific workflow of this mechanism is combined with... Figure 3The overall scanning and detection flowchart shown is explained in detail.

[0060] During the initial detection phase, the central control host computer 9 first drives the two-degree-of-freedom high-precision displacement stage 7 to move to the preset full-field scanning starting coordinate point P(x,y). The system determines whether this is the first detection start: if it is the first start, the analysis and control module controls the spatial light modulator 4 to load the default initial phase hologram, generate the initial excitation array, and excite Lamb waves on the sample surface. The laser Doppler vibrometer 8 collects the original ultrasonic signal at this point, and then the system executes... Figure 2 The complete inversion analysis process shown is as follows: physical information neural network mode decoupling and film thickness inversion, modal feature evaluation and optimal mode selection, target wavenumber extraction and hologram reconstruction. Finally, the optimal excitation parameters of the current detection point are determined and the high signal-to-noise ratio signal recording of the detection point is completed using these parameters.

[0061] After acquiring the signal at the current detection point, the system enters the scanning traversal logic. The central control host computer 9 controls the two-degree-of-freedom high-precision displacement stage 7 to move the probe head to the next detection coordinate point according to the preset scanning array 12 trajectory. At this time, the system does not re-trigger the complete initial inversion process at each new detection point, but adopts an efficient strategy of "prior parameter inheritance + adaptive evaluation and correction". Specifically, the system first uses the optimal excitation parameters determined at the previous detection point to quickly acquire the ultrasonic signal at the current detection point. Subsequently, the analysis and control module extracts and evaluates the feature parameters of the current signal. The feature parameters used for evaluation include, but are not limited to: modal purity, signal-to-noise ratio, and output residuals in the physical information neural network inversion process. If all of the above indicators are better than the preset thresholds, it indicates that the local physical characteristics of the current detection point—mainly the film thickness parameter—are basically consistent with those of the previous detection point, and the optimal excitation parameters of the previous detection point are still applicable at the current point. At this time, the system directly records the detection result of the current detection point and continues to jump to the next detection point. This strategy maximizes the reuse of existing optimization results and significantly improves the overall efficiency of the full-field scanning.

[0062] However, when faced with coatings of varying thickness and heterogeneity caused by fluctuations in the manufacturing process or non-uniform degradation during service, the physical properties such as local film thickness between adjacent detection points may change significantly. When this physical shift occurs, the signal at the current measurement point, obtained using the excitation parameters of the previous measurement point, will deteriorate in characteristic parameters such as modal purity, signal-to-noise ratio, or inversion residuals, manifesting as a deterioration in index values ​​and a drop below the preset threshold. Once the analysis and control module detects this deterioration, the central control host computer 9 automatically triggers an online closed-loop optimization program. This program requires no manual intervention and automatically returns to execute the complete inversion analysis process: using the newly acquired original signal at the current measurement point as input, it re-runs the physical information neural network to invert film thickness parameters, re-evaluates each modal component and locks the optimal mode at the current measurement point, recalculates the optimal wavenumber k and generates a new phase hologram, and finally drives the spatial light modulator 4 to refresh the hologram to reconstruct an excitation array that matches the local physical properties of the current measurement point. Through this timely online parameter correction, the system can quickly restore the signal quality at the current measurement point, bringing it back to the optimal excitation state. This closed-loop iterative mechanism ensures that every coordinate point in the scanning and detection process, regardless of the fluctuations in its local physical properties, can obtain the optimal excitation conditions that match its physical characteristics, thereby guaranteeing a high degree of consistency in key indicators such as signal-to-noise ratio and modal purity of the full-field scanning data.

[0063] After the two-degree-of-freedom high-precision displacement stage 7 has traversed all preset scanning coordinate points within the test area, the full-field signal acquisition process ends. The central control host computer 9 summarizes the local film thickness parameter h and ultrasonic feature values ​​such as time of flight and phase velocity extracted from each detection point under its optimal excitation state. Through spatial interpolation and image reconstruction algorithms, it generates a global film thickness distribution map and an image of internal defect distribution of the sample under test. Since the data from each measurement point are high-quality signals acquired under optimal excitation state, the reconstructed image is superior to the traditional fixed-parameter full-field scanning scheme in terms of spatial resolution, contrast, and quantitative analysis accuracy.

[0064] The unified time reference link 6 provides a nanosecond-level precision timing synchronization foundation for the precise coordination of the aforementioned excitation and acquisition processes. This link provides a unified time reference signal to the trigger output of the ns-pulse laser 1 and the sampling trigger input of the laser Doppler vibrometer 8 via a synchronized clock source, ensuring that the excitation time of the laser pulse and the signal acquisition start time of the Doppler vibrometer are strictly aligned within an extremely short time. This strict time synchronization is crucial for the accurate analysis of the multimodal dispersion characteristics of Lamb waves. During full-field multi-point scanning, the unified time reference ensures a consistent physical definition of the zero moment of the acquired signals at each detection point, thereby guaranteeing the comparability of ultrasonic time-of-flight measurements between different detection points and overcoming the phase random error introduced by trigger jitter. This condition is a necessary physical prerequisite for subsequent analysis of Lamb wave dispersion characteristics, separation of modal components, and reconstruction of the Lamb wave field image using algorithms such as spatial cross-correlation.

[0065] Using the technical solution provided by this invention, experimental verification was conducted on thermal barrier coating samples with unknown thickness distributions. Experiments show that, without any prior knowledge about the coating thickness, after the system initially excites and acquires the original signal at a single detection point, it completes the inversion of film thickness parameters and the selection and locking of the optimal mode within sub-second time using a physical information neural network. After the spatial light modulator refreshes based on the calculated hologram, the signal purity of the target Lamb wave mode is significantly improved compared to the initial excitation state, and the signal-to-noise ratio is significantly improved. During the full-field scanning process, when the displacement stage moves the detection point to a transition region where the coating thickness changes significantly, the system successfully detects the degradation trend of the signal indicators and automatically triggers an online closed-loop optimization program, completing the re-optimization of excitation parameters within seconds and restoring signal quality. The final reconstructed full-field film thickness distribution image matches well with the reference thickness value obtained by destructive metallography, verifying the technical effectiveness and practical value of this invention in the non-destructive characterization of thin film / coating structures with unknown or variable thickness.

[0066] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various improvements and equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and principles thereof, and such improvements and equivalent substitutions should also be considered to fall within the protection scope of the claims of the present invention.

Claims

1. A laser-ultrasound inspection system with adaptive selection of excitation of Lamb wave modes, characterized in that, include: The excitation light control link includes a pulsed laser, a beam homogenization component, a spatial light modulator (SLM), and a 4f lens group. The spatial light modulator is controlled by a central control host computer. By performing pixel-scale phase modulation on the incident laser, a laser linear array excitation source with dynamically adjustable spatial period, linewidth, and line number is generated to excite a Lamb wave signal on the surface of the sample under test. The ultrasonic signal acquisition link includes a laser Doppler vibrometer and a multi-degree-of-freedom high-precision displacement stage. The laser Doppler vibrometer is used to capture the Lamb wave time-domain signal of each detection point on the surface of the sample and transmit it to the central control host computer. The multi-degree-of-freedom high-precision displacement stage is used to drive the detection point switching and synchronously feed back the position coordinate information. The central control host computer integrates a physical information-driven analysis and control module. The analysis and control module is configured to: perform mode decoupling and feature extraction on the Lamb wave time-domain signal acquired by the ultrasonic signal acquisition link based on the physical constraints established by the Rayleigh-Lamb dispersion equation, at least inversely derive the local film thickness parameters of the current detection point, and calculate the optimal wavenumber of the target detection mode accordingly, and then generate a phase hologram for refreshing the spatial light modulator, so that the spatial parameters of the laser linear array excitation source and the wavelength of the target detection mode achieve adaptive matching; A unified time reference link is used to provide a unified time reference for the laser excitation time of the excitation light modulation link and the signal acquisition time of the ultrasonic signal acquisition link, so as to ensure the phase resolution accuracy of the multimodal dispersion characteristics.

2. The laser-ultrasound inspection system of claim 1, wherein, The beam homogenization assembly includes a beam expander and a diffractive optical homogenizer. The diffractive optical homogenizer is disposed in the optical path before the spatial light modulator and is used to shape the incident pulsed laser beam into a flat-topped light spot with uniform intensity distribution in space, so as to eliminate clutter interference introduced by the Gaussian energy gradient.

3. Laser-ultrasound inspection system for the adaptive selection excitation of Lamb wave modes according to claim 1 or 2, characterized in that, The central control host computer is preloaded with a physical information neural network (PINN) trained with physical constraints. The PINN uses the Rayleigh-Lamb dispersion equation as the regularization physical constraint term of the loss function to decouple the multimodal aliased Lamb wave signal contained in the original time domain signal of the acquired single point into independent modal components, and outputs the film thickness parameters and modal decoupling information of the current detection point in real time.

4. The laser-ultrasound inspection system of claim 3, wherein, The physical information-driven analysis and control module in the central control host computer is further configured to: based on the modal decoupling information output by the PINN, quantitatively extract the modal purity index and signal-to-noise ratio index of each independent modal component, automatically select the optimal detection mode of the current detection point from the multimodal coexistence field through a weighted evaluation function, and extract the target wavenumber corresponding to the optimal detection mode.

5. The Lamb wave mode adaptive selection excitation laser ultrasonic detection system according to claim 4, characterized in that, The physical information-driven analysis and control module in the central control host computer reconstructs the phase hologram of the spatial light modulator in real time based on the target wavenumber of the selected optimal detection mode and the inverted local film thickness parameters. It dynamically adjusts the spatial period and linewidth of the laser line array excitation source so that the spatial period matches the wavelength of the target detection mode, thereby achieving physical coherent enhancement excitation of the mode.

6. The Lamb wave mode adaptive selective excitation laser ultrasonic detection system according to claim 1, characterized in that, The central control host computer, the spatial light modulator, and the laser Doppler vibrometer form an adaptive closed-loop control link, which is equipped with online iterative optimization logic. During the full-field scanning detection process of the multi-degree-of-freedom high-precision displacement stage, when the signal characteristic parameters of the current detection point are detected to be lower than the preset threshold, the central control host computer automatically activates the parameter optimization program and re-executes the film thickness inversion, optimal mode locking and phase hologram refresh based on the ultrasonic signal of the current detection point until the signal characteristic parameters of the current detection point are restored to above the preset threshold.

7. The Lamb wave mode adaptive selective excitation laser ultrasonic detection system according to claim 6, characterized in that, The signal characteristic parameters include at least one of modal purity, signal-to-noise ratio, and inversion residual.

8. The Lamb wave mode adaptive selective excitation laser ultrasonic detection system according to claim 1, characterized in that, The multi-degree-of-freedom high-precision displacement stage is a two-degree-of-freedom high-precision displacement stage with a repeatability accuracy better than 0.5μm.

9. A laser ultrasonic detection method with adaptive selection of Lamb wave modes for excitation, implemented based on the system described in any one of claims 1 to 8, characterized in that, Includes the following steps: S1: Drive the multi-degree-of-freedom high-precision displacement stage to move to the current detection coordinates. The central control host computer drives the spatial light modulator to generate an initial linear array hologram, forming a laser linear array excitation source on the surface of the sample to be tested, and obtaining the original Lamb wave time domain signal of the current detection point through a laser Doppler vibrometer. S2: Using the analysis and control module driven by the pre-loaded physical information in the central control host computer, based on the physical constraints established by the Rayleigh-Lamb dispersion equation, the original Lamb wave time domain signal is decoupled in mode, the local film thickness parameters of the current detection point are inverted, and the optimal detection mode of the current detection point is selected from the decoupled multi-mode components in combination with the signal characteristic parameters. S3: Extract the target wavenumber corresponding to the optimal detection mode, reconstruct the phase hologram in real time and refresh the spatial light modulator, dynamically adjust the spatial parameters of the laser line array excitation source to match the wavelength of the target detection mode, realize the directional enhancement excitation of the mode, and complete the signal recording of the current detection point; S4: Drive the multi-degree-of-freedom high-precision displacement stage to switch to the next detection point. The system uses the optimal excitation parameters of the previous detection point to obtain the signal of the current detection point and executes adaptive branch logic judgment: If the signal characteristic parameters and inversion residuals of the current detection point are better than the preset threshold, the detection result is recorded directly and the process jumps to step S5. If the signal characteristic parameters or inversion residuals at the current detection point are worse than the preset threshold, the closed-loop optimization procedure will be automatically triggered, and the process will return to step S2 to re-execute the film thickness inversion, mode locking, and phase hologram refresh. S5: Traverse all detection points within the test area, summarize the local film thickness parameters and ultrasonic feature values ​​extracted from each detection point under the optimal excitation state, and reconstruct the global structural distribution of the sample under test.

10. The laser ultrasonic detection method with adaptive selective excitation of Lamb wave modes according to claim 9, characterized in that, Step S2, which utilizes the pre-loaded physical information in the central control host computer to drive the analysis and control module to perform modal decoupling and film thickness inversion, specifically includes: The acquired raw Lamb wave time-domain signal is mapped to a two-dimensional time-frequency plane through time-frequency transformation; The pre-trained physical information neural network is invoked, and the Rayleigh-Lamb dispersion equation is used as the physical constraint regularization term of the loss function to search for the global optimal solution that minimizes the residual in the solution space. At the same time, the decoupled independent modal components and the corresponding local film thickness parameters are output.

Images