Material detection method and system based on photoacoustic effect

Abstract

The invention relates to the technical field of material detection, and provides a material detection method and system based on a photoacoustic effect. The material detection system comprises a light source for sending first detection light; the dispersion assembly is used for receiving the first detection light and obtaining second detection light after dispersion; the condensation assembly is used for focusing the second detection light on the surface of the to-be-detected material; the acquisition assembly is used for collecting acoustic signals; and the processor determines detection parameters of the to-be-detected material based on the sound wave signal, wherein the detection parameters comprise one or more of a material, a surface process, impurities or damage. According to the material detection system provided by the invention, the multi-wavelength photoacoustic response spectrum of the material can be captured by realizing single excitation through continuous wavelength, and the technical problem that a traditional detection system is easily interfered by environmental noise and material surface states due to adoption of a single signal is solved, so that the detection precision is effectively ensured, detection parameters are multi-dimensionalized, and the detection accuracy is improved. Various parameters of the to-be-detected material can be obtained, and the detection efficiency is improved.

Description

Technical Field

[0001] This application relates to the field of materials testing technology, and in particular to a materials testing method and system based on photoacoustic effect. Background Technology

[0002] Photoacoustic spectroscopy, as a powerful means of material analysis, has been successfully applied in fields such as gas detection and biomedical imaging due to its inherent advantages, such as direct detection of optical absorption, insensitivity to samples with strong scattering, and high sensitivity.

[0003] However, when photoacoustic technology is applied to the surface and composition analysis of solid materials, the following problems exist. First, the traditional method of constructing absorption spectra through point-by-point wavelength scanning is time-consuming and cannot meet the requirements of high throughput and real-time performance. Second, single-wavelength lasers are typically used, primarily to excite ultrasonic echoes for assessing geometric and mechanical defects (such as cracks and delamination), lacking the ability to specifically identify the chemical composition of materials. Even with wavelength scanning, the information obtained focuses mainly on optical absorption characteristics, resulting in insufficient discrimination (specificity) for materials with similar optical absorption but different mechanical or thermal properties. These problems lead to low detection efficiency and a limited range of detection parameters. Summary of the Invention

[0004] This application provides a material detection method and system based on photoacoustic effect to solve the technical problems of low detection efficiency and limited detection parameter types in existing detection methods.

[0005] The first aspect of this application provides a material detection system based on photoacoustic effect, comprising: a light source configured to emit a first detection light; the first detection light including a continuous wavelength; a dispersion component disposed in the optical path of the first detection light, configured to receive the first detection light and disperse it to obtain a second detection light; the second detection light including a preset wavelength, the preset wavelength being any wavelength among the continuous wavelengths; a focusing component disposed in the optical path of the second detection light, configured to focus the second detection light onto the surface of the material to be detected; a collection component disposed between the focusing component and the material to be detected, configured to collect acoustic signals generated by the second detection light acting on the material to be detected based on the photoacoustic effect; and a processor connected to the collection component, configured to determine detection parameters of the material to be detected based on the acoustic signals, the detection parameters including one or more of material, surface process, impurities, or damage.

[0006] In some feasible implementations, the light source is a supercontinuum femtosecond fiber laser with a continuous wavelength range of 0.4-2.5 μm.

[0007] In some feasible implementations, the dispersive component includes: a first grating, a second grating, and a slit stop; the first grating and the second grating are arranged opposite to and parallel to each other, the first grating is configured to propagate a first detection light to the second grating; the second grating is configured to propagate the first detection light to the slit stop; the slit stop is configured to filter out a second detection light from the first detection light; wherein the slit stop is configured to be movable relative to the second grating.

[0008] In some feasible implementations, the focusing assembly includes a first galvanometer, a second galvanometer, and a field lens; the first galvanometer is disposed in the optical path of the second detection light and is configured to propagate the second detection light to the second galvanometer; the second galvanometer is configured to propagate the second detection light to the field lens; and the field lens is configured to focus the second detection light onto the surface of the material to be detected.

[0009] In some feasible implementations, the acquisition component includes multiple acquisition units arranged in an array; the acquisition units are configured to receive acoustic signals and send the acoustic signals to the processor.

[0010] In some feasible implementations, the length of the slit stop is 0.15-1 mm.

[0011] In some feasible implementations, the focusing assembly also includes a driver connected to the first and second galvanometers respectively; the driver is configured to control the oscillation of the first and second galvanometers to adjust the propagation direction of the second detection light.

[0012] In some feasible implementations, the processor includes an acquisition module, an establishment module, a training module, and a recognition module; the acquisition module is configured to acquire training data; the establishment module is configured to establish a recognition model; the training module is configured to train the recognition model using the training data; the input of the recognition model is an acoustic signal, and the output of the recognition model is the detection parameters of the material to be detected; the recognition module is configured to use the recognition model to identify the detection parameters of the material to be detected.

[0013] In some feasible implementations, the training data includes a training set, a validation set, and a test set; the training set is configured to train the recognition model; the validation set is configured to validate the recognition model; and the test set is configured to test the recognition model; wherein the ratio of the training set, validation set, and test set is 4:3:3.

[0014] The photoacoustic material detection system provided in this application can capture the multi-wavelength photoacoustic response spectrum of a material with a single excitation using continuous wavelengths, rather than a single signal. This solves the technical problem of traditional detection systems being susceptible to interference from environmental noise and material surface conditions caused by the single signal used, thus effectively ensuring detection accuracy. Furthermore, the detection parameters are multidimensional, allowing the acquisition of multiple parameters of the material to be detected, thereby improving detection efficiency.

[0015] The second aspect of this application provides a material detection method based on photoacoustic effect, comprising: sending a first detection light; receiving the first detection light and dispersing it to obtain a second detection light; focusing the second detection light on the surface of the material to be tested; collecting an acoustic signal; the acoustic signal is generated by the second detection light acting on the material to be tested; determining detection parameters of the material to be tested based on the acoustic signal; the detection parameters include one or more of material, surface process, impurities or damage.

[0016] The photoacoustic effect-based material detection method provided in the second aspect of this application is applied to a photoacoustic effect-based material detection system. Its beneficial technical effects can be found in the first aspect, and will not be repeated here. Attached Figure Description

[0017] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the structure of a material detection system based on photoacoustic effect provided in an embodiment of this application; Figure 2 This is a structural block diagram of a processor provided in an embodiment of this application; Figure 3 This is a schematic diagram of the wavelength-ultrasonic intensity spectrum of the two alloys provided in the embodiments of this application; Figure 4 This is a schematic diagram of the wavelength-ultrasonic intensity spectrum of two alloys from different batches provided in the embodiments of this application; Figure 5 This is a schematic flowchart of a material detection method based on photoacoustic effect provided in an embodiment of this application.

[0019] Illustration markings: 1-Material testing system; 10-Light source; 20-Dispersion component; 21-First grating; 22-Second grating; 23-Slit aperture; 30-Concentrating component; 31-First galvanometer; 32-Second galvanometer; 33-Field lens; 34-Driver; 40-Acquisition component; 41-Acquisition device; 50-Processor; 51-Acquisition module; 52-Establishment module; 53-Training module; 54-Identification module; 2-Material to be tested. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are all within the protection scope of this application.

[0021] In the following description, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0022] Furthermore, in this application, directional terms such as "upper," "lower," "inner," and "outer" are defined relative to the indicated placement of the components in the accompanying drawings. It should be understood that these directional terms are relative concepts, used for relative description and clarification, and can change accordingly depending on the placement of the components in the accompanying drawings.

[0023] To facilitate the explanation of the technical solution of this application, some concepts involved in this application will be explained first below.

[0024] Photoacoustic effect: After a material absorbs light energy, the absorption area undergoes instantaneous thermal expansion or contraction due to the conversion of energy into heat energy, thereby generating elastic sound waves. This is the core principle of technologies such as photoacoustic spectroscopy.

[0025] Acoustic fingerprint: The photoacoustic signal spectrum (such as the "wavelength-ultrasonic intensity" spectrum) generated by a material under photoexcitation is unique due to the differences in absorption characteristics of different materials, and can be used as a specific feature to distinguish and identify the composition of a material.

[0026] Photoacoustic spectroscopy (PAS) is a powerful material analysis technique based on the photoacoustic effect: when a substance absorbs modulated light energy, it undergoes thermoelastic expansion, exciting ultrasonic waves. This technique infers the composition and properties of a substance by "listening" to the photoinduced "sound." Due to its inherent advantages such as directly detecting optical absorption, insensitivity to strongly scattering samples, and high sensitivity, it has been successfully applied in fields such as gas detection and biomedical imaging.

[0027] However, when photoacoustic technology is applied to the surface and composition analysis of solid materials (especially metal alloys and composite materials), traditional technical solutions reveal several core bottlenecks that urgently need to be addressed, severely restricting their widespread application: ① Traditional solutions rely on point-by-point wavelength scanning with tunable lasers to construct absorption spectra. This serial approach results in lengthy data acquisition times, failing to meet the stringent requirements of high throughput and real-time performance for industrial online inspection. ② Most traditional systems use only single-wavelength lasers, primarily exciting ultrasonic echoes used to assess geometric and mechanical defects (such as cracks and delamination), lacking the ability to specifically identify the chemical composition of materials. Even with wavelength scanning, the information obtained focuses mainly on optical absorption characteristics, resulting in insufficient differentiation (specificity) for materials with similar optical absorption but different mechanical or thermal properties. These problems lead to low detection efficiency and a limited range of detection parameters.

[0028] To address the aforementioned technical issues, this application provides a photoacoustic-based material detection system. By integrating optics, supercontinuum lasers, microphone arrays, and deep learning algorithms, and organically combining these advanced cross-disciplinary technologies, a material detection system capable of high-speed, high-specificity, and high-robustness identification and quantification of solid material composition is constructed. This enables the rapid, non-destructive, and highly specific online identification and quantification analysis of the surface and near-surface composition of a wide range of solid materials across multiple categories.

[0029] Figure 1 This is a schematic diagram of the structure of a material detection system 1 based on photoacoustic effect provided in an embodiment of this application.

[0030] See Figure 1 As shown, the photoacoustic material detection system 1 provided in this application includes a light source 10, a dispersion component 20, a focusing component 30, a collection component 40, and a processor 50.

[0031] The light source 10 is used to transmit a first detection light, wherein the first detection light is a detection light of continuous wavelength. That is, the first detection light includes detection light of multiple wavelengths.

[0032] A dispersion component 20 is disposed in the optical path of the first detection light. The dispersion component 20 is used to receive the first detection light and disperse it to obtain a second detection light. The second detection light includes a preset wavelength, and the preset wavelength is any one of the continuous wavelengths. That is to say, the first detection light, which has multiple wavelengths, can be separated into a second detection light of the preset wavelength after passing through the dispersion component 20.

[0033] The focusing component 30 is disposed in the optical path of the second detection light. The focusing component 30 is used to focus the second detection light onto the surface of the material to be tested 2 for non-destructive testing of the material to be tested 2.

[0034] The acquisition component 40 is positioned between the focusing component 30 and the material to be tested 2. After the second detection light is focused on the surface of the material to be tested 2, an acoustic signal is generated due to the photoacoustic effect. The acoustic signal is an ultrasonic signal. The acoustic signal propagates inside the object. Thus, not only the surface characteristics of the material to be tested 2 can be determined by the acoustic signal, but also the internal characteristics of the material to be tested 2.

[0035] The processor 50 is connected to the acquisition component 40. The processor 50 is used to receive the acoustic wave signal sent by the acquisition component 40 and determine the detection parameters of the material 2 to be tested based on the acoustic wave signal. The detection parameters may include one or more of the following: material, surface process, impurities or damage.

[0036] For example, when the detection parameters include material, the processor 50 can determine the material of the material to be detected 2 as iron, aluminum, or steel based on the acoustic signal. Specifically, when the material to be detected 2 is steel, the processor 50 can determine that the material to be detected 2 is 304 stainless steel or 316 stainless steel based on the acoustic signal.

[0037] When the detection parameters include surface process, the processor 50 can determine the surface process of the material 2 to be detected as polishing, coating, oxidation or sandblasting based on the acoustic signal.

[0038] When the detection parameters include impurities, the processor 50 can determine whether the surface or near-surface of the material 2 to be detected contains impurities, such as sulfur impurities in steel, based on the acoustic signal. Specifically, the content of impurities can also be determined by the acoustic signal, such as 1% sulfur impurities in steel.

[0039] When the detection parameters include damage, the processor 50 can determine, based on the acoustic signal, that the surface or interior of the material 2 to be detected contains damage. The damage can include surface damage and internal damage, and the specific location of the damage can be determined by the acoustic signal.

[0040] It is worth noting that the detection parameters can include one or more. When there are multiple detection parameters, multiple detection parameters of the material 2 to be tested can be detected simultaneously to improve the detection accuracy.

[0041] The photoacoustic material detection system 1 provided in this application embodiment can capture the multi-wavelength photoacoustic response spectrum of a material with a single excitation using continuous wavelengths, rather than a single signal. This solves the technical problem of traditional detection systems being susceptible to interference from environmental noise and material surface conditions due to the use of a single wavelength. It has the ability to specifically identify the chemical composition of materials and can achieve rapid detection, meeting the requirements of high throughput and real-time performance. This effectively ensures detection accuracy, and the detection parameters are multidimensional, allowing the acquisition of multiple parameters of the material 2 to be detected, thereby improving detection efficiency.

[0042] In some feasible implementations, the light source 10 is a supercontinuum femtosecond fiber laser with a continuous wavelength range of 0.4-2.5 μm.

[0043] Understandably, traditional photoacoustic detection often relies on fixed single-wavelength detection lasers, such as 808nm or 1064nm, which depend on low-dimensional features like "single-wavelength acoustic signal amplitude." This can only excite a limited number of characteristic absorption peaks in the material being tested, making it susceptible to interference from environmental noise and material surface conditions. In contrast, this application uses a 400-2500nm supercontinuum laser. Targeting the absorption differences of different materials (such as plastics, metals, and biological tissues) at specific wavelengths in the supercontinuum, it can simultaneously match the material's characteristic absorption peaks across the entire wavelength range from the visible light absorption region to the near-infrared II region. A single excitation can capture the material's multi-wavelength photoacoustic response spectrum, rather than a single signal.

[0044] In some feasible implementations, the pulse width of the light source 10 is 200 fs; the repetition frequency is 50 kHz to 1 MHz; the output power is 5 W; and the spectral flatness is ±3 dB.

[0045] In this way, the femtosecond pulse duration is short, only 50~200 fs, and the excitation is completed before heat diffusion. Therefore, the thermal diffusion effect can be ignored, avoiding the material surface burns and structural damage caused by heat accumulation in traditional nanosecond lasers. It is especially suitable for non-destructive testing of fragile materials (such as biological tissues and precision electronic components). It not only improves the detection range of the type of material to be tested, but also reduces the risk of damage to the material to be tested during the testing process.

[0046] In some feasible implementations, the peak power of the femtosecond pulse of the light source 10 can be 10. 6 -10 9 W. Thus, at this pulse peak power, weak photoacoustic signals from low-concentration impurities (such as 0.1% harmful additives in the material) and subsurface structures (such as the substrate components under the coating) can be excited; the near-infrared II region (1100~2500nm) has strong light penetration, and the detection depth of biological tissues and polymer materials reaches the millimeter level, breaking through the traditional limitation of "only being able to detect surface / high concentration components", and has good weak signal detection capability.

[0047] See also Figure 1 As shown, in some feasible implementations, the dispersive component 20 may include a first grating 21, a second grating 22, and a slit stop 23.

[0048] The first grating 21 and the second grating 22 are arranged opposite to each other and parallel to each other. A first detection light is incident on the first grating 21 along a first direction, and diffraction occurs on the first grating 21. The first grating 21 is used to change the propagation direction of the first detection light, propagating it along a second direction to the second grating 22. The first detection light also diffracts on the second grating 22, and the second grating 22 is used to change the propagation direction of the first detection light back from the second direction to the first direction, allowing it to propagate to the slit stop 23.

[0049] Specifically, the first grating 21 and the second grating 22 can form a grating pair structure. By setting the grating pair, the spatial dispersion of supercontinuum laser can be achieved by utilizing the dispersion capability of the grating pair while keeping the beam direction of the first detection light unchanged.

[0050] The slit stop 23 is configured to filter out the second detection light from the first detection light; wherein the slit stop 23 is configured to be movable relative to the second grating 22.

[0051] Specifically, the position of the slit aperture 23 can correspond to the wavelength of the second detection light. By adjusting the position of the slit aperture 23, the second detection light of a preset wavelength in the first detection light can pass through the slit aperture 23. In other words, by matching slit apertures 23 with different positions and widths, the second detection light of a preset wavelength can be filtered out.

[0052] In some feasible implementations, the first grating 21 and the second grating 22 have the same line density, both being 1700 lines / mm to 1900 lines / mm. For example, the line density of the first grating 21 and the second grating 22 can be one of 1700 lines / mm, 1800 lines / mm, or 1900 lines / mm. Of course, the line density of the first grating 21 and the second grating 22 can also be other values ​​within the range of 1700 lines / mm to 1900 lines / mm.

[0053] By setting the line density of the first grating 21 and the second grating 22 in the range of 1700 lines / mm to 1900 lines / mm, the first grating 21 and the second grating 22 have high dispersion capability and resolution.

[0054] In some feasible implementations, the length of the slit stop 23 is adjustable in the range of 0.15-1mm, and the corresponding resolution is 1-0.66nm.

[0055] Specifically, when the length of the slit stop 23 is 0.15 mm, the corresponding resolution is 1 nm. When the length of the slit stop 23 is 1 mm, the corresponding resolution is 0.66 nm. In this way, the resolution can be adjusted by changing the size of the slit stop 23 to match the required resolution.

[0056] See also the following for some feasible implementation methods. Figure 1 As shown, the focusing assembly 30 may include a first galvanometer 31, a second galvanometer 32, and a field mirror 33.

[0057] The first galvanometer 31 is disposed in the optical path of the second detection light. The first galvanometer 31 is used to change the propagation direction of the second detection light and propagate the second detection light to the second galvanometer 32. The second galvanometer 32 is used to change the propagation direction of the second detection light and propagate the second detection light to the field mirror 33. The field mirror 33 is used to focus the second detection light onto the surface of the material 2 to be detected.

[0058] By setting the first galvanometer 31 and the second galvanometer 32, the direction of the second detection light can be adjusted. By setting the field lens 33, the second detection light can be focused on the surface of the material 2 to be detected.

[0059] During the focusing process of the second detection light, it is necessary to adjust the angles between the first galvanometer 31 and the second galvanometer 32 and the horizontal plane. The focusing assembly 30 may also include a driving member 34, which is connected to the first galvanometer 31 and the second galvanometer 32 respectively, and is used to control the swing of the first galvanometer 31 and the second galvanometer 32 to adjust the output direction of the second detection light for easier focusing.

[0060] It should be emphasized that the number of driving components 34 can be two or one. When there are two driving components 34, the two driving components 34 separately control the first galvanometer 31 and the second galvanometer 32; or when there is one driving component 34, one driving component 34 simultaneously drives the first galvanometer 31 and the second galvanometer 32. The driving component 34 is connected to a power source, which provides power to the driving component 34. For example, the driving component 34 can be a drive motor.

[0061] See also the following for some feasible implementation methods. Figure 1 As shown, the acquisition component 40 may include multiple acquisition units 41, which are arranged in an array and a preset distance is provided between the multiple acquisition units 41 and the material to be tested 2, for acquiring sound wave signals from multiple directions and sending the acquired sound wave signals to the processor 50.

[0062] By arranging multiple collectors 41 in an array, it is possible to receive acoustic signals that penetrate the interior of the material 2 under test, enabling large-area, real-time, online monitoring.

[0063] Specifically, the acquisition unit 41 can be a sensitive capacitive micro-electro-mechanical system (MEMS), and the acquisition component 40 can be set as a 4-element micro planar array. The frequency response is 10kHz-1MHz, the sampling rate is ≥2MHz, and the synchronization delay is in the microsecond range; the sensitivity is ≤-35dBV / Pa, and the SNR is ≥70dB. In this way, the 4-element MEMS microphone array focuses the laser excitation point through beamforming, with a beamwidth of 30°-60°. This filters out environmental noise such as the driving components 34 of the first and second galvanometers 31 and the laser heat dissipation, improving the signal-to-noise ratio to over 70dB; the weak signal detection limit is reduced by more than 10 times, enabling the identification of trace impurities as low as 0.01% in the material 2 under test. It exhibits strong anti-interference capabilities and a low weak signal detection limit, effectively improving the detection efficiency of the material 2 under test.

[0064] It is important to emphasize that this application employs a 400-2500 nm supercontinuum laser. The supercontinuum covers the absorption peaks of the material across the entire visible and near-infrared spectrum, forming a wavelength-ultrasonic intensity spectral line, much like an acoustic fingerprint. The spectral line morphology varies significantly among different materials, fundamentally avoiding misjudgments caused by overlapping features in traditional techniques and effectively improving the specificity of component resolution. For example, the captured "wavelength (400~2500 nm) - acoustic signal intensity" spectral line is material-specific (similar to an "acoustic fingerprint"). For instance, polyethylene has a strong absorption peak at 1720 nm, corresponding to a sharp increase in acoustic signal amplitude; while polypropylene has an absorption peak at 1150 nm, resulting in completely different spectral line morphologies for the acoustic signal, thus enhancing the specificity of component resolution from the root of the characteristic features.

[0065] The processor 50 is connected to the data acquisition unit 41. The processor 50 is used to determine the detection parameters of the material 2 to be tested based on the acoustic wave signal. The detection parameters include one or more of the following: material, surface finish, impurities, or damage. Thus, after receiving the acoustic wave signal sent by the data acquisition unit 41, the processor 50 can determine the detection parameters of the material 2 to be tested based on the acoustic wave signal, such as the detection result being 304 stainless steel with surface damage, or the detection result being 304 stainless steel containing 1% impurities.

[0066] The processor 50 can be connected to the collector 41 via wired or wireless connection.

[0067] Figure 2 This is a structural block diagram of a processor provided in an embodiment of this application.

[0068] Among some feasible implementation methods, see Figure 2 As shown, the processor 50 may include an acquisition module 51, an establishment module 52, a training module 53, and a recognition module 54.

[0069] The acquisition module 51 is used to acquire training data. To ensure data diversity, acquiring training data can be achieved through data collection and data preprocessing.

[0070] During data collection, the data needs to be labeled to ensure both diversity and consistency. Data diversity aims to cover the material being tested, while consistency aims to avoid irrelevant interference.

[0071] To effectively identify the material differences and surface process variations of the tested material 2 using data and avoid confusion, the following sampling methods were employed during data collection: common metals, plastics, ceramics, and other materials were tested; different surface processes of the same material were sampled, such as polishing, coating, oxidation, and sandblasting; and different component proportions of the same material were sampled, such as 5052 aluminum alloy, 6061 aluminum alloy, and 7075 aluminum alloy.

[0072] During data acquisition, experimental variables such as the power of the first detection light and the position of the microphone array are controlled. Five sets of repeated data are collected and normalized; comparison of these sets eliminates irrelevant interference and improves the database's accuracy. The training data also includes combinations of different parameters, such as "material-surface process" combinations and "material-damage" combinations. Furthermore, for each combination, at least 100 sets of valid data are collected to avoid overfitting.

[0073] After data collection is complete, data preprocessing is required. Data preprocessing can be achieved through noise reduction, normalization, and outlier removal.

[0074] In some feasible implementations, the denoising process can include two parts: the first part can target the acquisition unit 41, and the second part can target the optical path. In the first part, for low-frequency or random noise in the environment where the acquisition unit 41 is located, wavelet transform can be used to denoise while preserving parameters such as spectral peaks, or moving average filtering can be used to smooth high-frequency noise, with a window size of 5-10 wavelet lengths. In the second part, for optical path drift, such as an overall drift in the intensity of a certain wavelength, baseline correction methods can be used, such as subtracting the baseline after polynomial fitting to eliminate background interference.

[0075] The purpose of normalization is to eliminate intensity differences in different measurements of the same type of material caused by laser power fluctuations. For example, when the laser power is high, the intensity of all wavelengths is higher, but the spectral line shape in the acoustic signal remains unchanged. During normalization, Min-Max normalization can be used, which makes the spectral line shape in the acoustic signal more intuitive and suitable for subsequent feature extraction.

[0076] The purpose of outlier removal is to avoid outlier data interfering with the training model. The 3σ rule can be used. For example, if the intensity of a sound wave signal of a certain wavelength deviates from the mean of all samples of the material by more than 3 times the standard deviation, it is judged as an outlier, such as when the acquisition device 41 is blocked during measurement.

[0077] Module 52 is used to establish a recognition model, which is used to identify the detection parameters of the material to be tested.

[0078] In some feasible implementations, the recognition model can be a 1D convolutional neural network (1D-CNN) deep learning model or a Transformer deep learning model. The recognition model can include interconnected convolutional layers, pooling layers, and fully connected layers. Convolutional layers are used to extract local features from the acoustic signal, such as peaks, shape, and full width at half maximum (FWHM). Pooling layers are used to preserve key features. Fully connected layers are used for classification.

[0079] The training module 53 is used to train the recognition model using training data; the input of the recognition model is the acoustic signal, and the output of the recognition model is the detection parameters of the material to be detected 2, which include one or more of the following: material, surface process, impurities or damage.

[0080] In some feasible implementations, during the training process, the training data can be assigned a type, which may include a training set, a validation set, and a test set.

[0081] The training set is used to train the recognition model, enabling it to learn the inherent patterns and regularities of the data aggregation; the validation set is used to validate the recognition model and guide its development; the test set is used to test the recognition model and evaluate its generalization ability.

[0082] In some feasible implementations, the ratio of training set, validation set, and test set is 4:3:3. In this implementation, the data can be stratified according to material category, for example, 40% training set, 30% validation set, and 30% test set. It is important to emphasize that during type matching, the data cannot be randomly allocated. For example, out of 100 data sets for the same material, 40 sets might be in the training set and 30 sets in the test set. The data should be allocated according to sample batches; for example, all material data collected in a certain batch could be included in the test set to simulate the real-world scenario of "unknown new samples" and avoid data leakage.

[0083] In some feasible implementations, the loss function during training can be adjusted based on the relationship between sample sizes. For example, if the sample sizes of each class are similar, the cross-entropy loss function is chosen; if the classes are imbalanced, the weighted cross-entropy loss function is used. The optimizer prioritizes Adam learning rate scheduling and uses cosine annealing scheduling (CAS). Regularization techniques can include SVM with L2 regularization, Dropout layers, batch normalization, and weight decay. To prevent overfitting, early stopping can be implemented, including monitoring the accuracy on the validation set. If the accuracy does not improve (or the loss increases) for 5-10 consecutive epochs, training is stopped, and the current model is saved as the optimal model.

[0084] In some feasible implementations, model evaluation can be performed after training is complete. Model evaluation can be achieved through accuracy, recall, F1 score, confusion matrix, etc.

[0085] After the recognition model is trained, it needs to be linked with the optical path and the acoustic signal.

[0086] Specifically, the decision-making basis of the SHAP value or LIME analysis and identification module 54 can be adopted. If the identification model determines that the core feature of "stainless steel-polished" is the high intensity response of the 650nm wavelength acoustic signal, the center wavelength of the slit aperture 23 can be fixed near 650nm to reduce invalid measurements of other wavelengths and improve the detection speed.

[0087] In some feasible implementations, to improve the recognition model's adaptability to different material types, incremental training can be performed to adapt to unknown material types. Specifically, when a new material type is added, such as introducing "titanium alloy," it is not necessary to retrain the entire recognition model; incremental training reduces data requirements and training time. In incremental training, the weights added to the recognition model only train the parameters corresponding to the new material type. These parameters can represent the differences between the new material type and known materials, such as hardness.

[0088] In some feasible implementations, the data from the recognition model can be exported in the Open Neural Network Exchange (ONNX) format. The ONNX format is cross-platform compatible. The ONNX format data can then be embedded into the software integrated into the processor 50, such as a host computer written in C++ / Python, to achieve real-time acquisition of the acoustic signal spectrum, real-time detection, and real-time result display.

[0089] In this implementation, a "human error correction" module can be added to the recognition model. When the operator finds that the recognition module 54 has made a misjudgment, the correct label is marked and sent back to the processor 50. Within a preset period, the model is fine-tuned and trained using newly accumulated labeled data to continuously improve its accuracy. The preset period may be every half month, every month, or every three months.

[0090] The recognition module 54 is used to identify the detection parameters of the material 2 to be tested using the recognition model. Specifically, after the recognition model has been trained and the test results meet the preset requirements, the detection parameters of the material 2 to be tested can be detected.

[0091] It is important to emphasize that traditional detection methods, such as spectroscopic component analysis, require manual comparison with standard spectral libraries, rely on professional experience, and struggle to handle complex scenarios involving multi-component mixtures and overlapping spectral lines. The embodiments of this application directly input high-dimensional photoacoustic spectral features (hundreds to thousands of dimensions) into a recognition model (such as CNN or Transformer). This model automatically learns the nonlinear mapping relationship between "high-dimensional photoacoustic spectral morphology - material composition - content percentage," making it an AI model that eliminates the need for manual feature extraction. The recognition model offers high efficiency in component analysis, achieving integrated qualitative and quantitative analysis.

[0092] In qualitative identification, single-point component identification can be completed quickly, significantly improving efficiency compared to traditional manual comparison. In quantitative identification, the proportion of each component can be output simultaneously for mixed materials (such as plastic alloys and compound drugs), overcoming the bottleneck of easy qualitative identification but difficult quantitative identification. Addressing the data heterogeneity caused by large-format scanning, such as "edge signal attenuation and environmental fluctuations," the identification model can achieve "point-by-point feature correction" through transfer learning and spatial attention mechanisms, and automatically stitch together the component distribution map of the entire region, ultimately outputting a two-dimensional component heatmap with "millimeter-level spatial resolution + percentage-level content accuracy," avoiding boundary errors caused by traditional stitching.

[0093] The photoacoustic material detection system 1 provided in this application embodiment is based on the photoacoustic effect generated by the interaction between light and materials. It utilizes the dispersion capabilities of the first grating 21 and the second grating 22 to achieve supercontinuum laser dispersion. A preset wavelength of the first detection light is filtered out according to the matching slit aperture 23. The filtered second detection light is then focused onto the surface of the material to be detected 2 through the first galvanometer 31, the second galvanometer 32, and the field mirror 33. An array of collectors 41 records the acoustic signal generated by the interaction between the second detection light and the material to be detected 2, forming an ultrasonic intensity spectrum of wavelength-acoustic signal. This spectrum is input into a trained recognition model, and the specific composition or surface treatment of the material to be detected 2 is identified by distinguishing the characteristics of the spectrum, thus achieving non-destructive, large-format material surface composition measurement.

[0094] To better understand the testing parameters of material 2 to be tested in this application, the following is a brief explanation. Figure 3 and Figure 4 A brief introduction to the ultrasonic intensity spectrum of the test results.

[0095] Figure 3 This is a schematic diagram of the wavelength-ultrasonic intensity spectrum of the two alloys provided in the embodiments of this application.

[0096] See Figure 3 As shown, the two alloys are 6061 aluminum alloy and 7075 aluminum alloy. The horizontal axis represents the wavelength of the ultrasonic wave propagating in the material under test 2, ranging from 500nm to 2500nm. The vertical axis represents the normalized ultrasonic intensity of the ultrasonic signal, ranging from approximately 0.00 to 1.1. Normalization is performed to eliminate interference from absolute values, facilitating a clear comparison of the trends and differences between the two curves.

[0097] Each curve shows the intensity attenuation of an acoustic signal at different wavelengths after passing through the corresponding alloy material. The curve for 6061 aluminum alloy has larger overall fluctuations. It exhibits high ultrasonic penetration at the peaks. The curve for 7075 aluminum alloy is relatively flatter, but its normalized intensity value is generally lower than that of 6061 aluminum alloy across the entire wavelength range, and the peaks of the two are different.

[0098] Figure 4 This is a schematic diagram of the wavelength-ultrasonic intensity spectrum of two alloys from different batches provided in the embodiments of this application. Among them, Figure 4 Figures (a) and (b) show schematic diagrams of wavelength-ultrasonic intensity spectra of 6061 aluminum alloy and 7075 aluminum alloy from the same batch, respectively. Figure 4 Figures (c) and (d) show schematic diagrams of wavelength-ultrasonic intensity spectra of 6061 aluminum alloy and 7075 aluminum alloy from another batch, respectively.

[0099] See Figure 4 As shown, the detection parameters are predicted using the photoacoustic effect-based material detection system 1 provided in this application embodiment. The prediction results match the actual results, with a confidence level of 1.

[0100] Specifically, Figure 4 In (a) and (c), the specific shapes of the two curves are not exactly the same, but their overall trends and characteristic patterns are similar, which reflects the normal data fluctuations of the same material under different conditions. Figure 4 The same applies to (b) and (c).

[0101] Corresponding to the aforementioned embodiment of the photoacoustic effect-based material detection system 1, this application also provides an embodiment of a photoacoustic effect-based material detection method. This material detection method is applied to the photoacoustic effect-based material detection system 1 provided in any of the above embodiments.

[0102] Figure 5 This is a schematic flowchart of a material detection method based on photoacoustic effect provided in an embodiment of this application.

[0103] See Figure 5 As shown, the material detection method based on photoacoustic effect can be implemented by the following steps S1 to S5.

[0104] Step S1: Send the first detection light.

[0105] In step S1, sending the first detection light can be achieved by the light source 10.

[0106] Step S2: Receive the first detection light, disperse it to obtain the second detection light.

[0107] In step S2, the dispersion component 20 receives the first detection light and disperses it to obtain the second detection light.

[0108] Step S3: Focus the second detection light onto the surface of the material to be tested 2.

[0109] In step S3, the focusing of the second detection light can be achieved by the focusing component 30.

[0110] Step S4: Collect acoustic signals; the acoustic signals are generated by the second detection light acting on the material to be detected 2.

[0111] In step S4, the collection of acoustic wave signals can be achieved by the acquisition component 40.

[0112] Step S5: Determine the detection parameters of the material to be tested 2 based on the acoustic signal; the detection parameters include one or more of the following: material, surface process, impurities or damage.

[0113] In step S5, the determination of material parameters can be achieved by the processor 50, specifically by the recognition model in the processor 50.

[0114] It should be noted that, upon considering the specification and practicing the application disclosed herein, those skilled in the art will readily conceive of other embodiments of this application. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.

[0115] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The true scope is indicated by this application.

Claims

1. A material detection system based on photoacoustic effect, characterized in that, include: The light source is configured to send a first detection light; The first detection light includes continuous wavelengths; A dispersion component is disposed in the optical path of the first detection light. The dispersion component is configured to receive the first detection light and disperse it to obtain a second detection light. The second detection light includes a preset wavelength, which is any wavelength among the continuous wavelengths. A focusing component is disposed in the optical path of the second detection light, and the focusing component is configured to focus the second detection light onto the surface of the material to be detected; A collection component is disposed between the light-concentrating component and the material to be tested. The collection component is configured to collect acoustic signals, which are generated by the second detection light acting on the material to be tested based on the photoacoustic effect. A processor, connected to the acquisition component, is configured to determine detection parameters of the material to be tested based on the acoustic signal, the detection parameters including one or more of material, surface finish, impurities, or damage.

2. The material detection system based on photoacoustic effect according to claim 1, characterized in that, The light source is a supercontinuum femtosecond fiber laser, and the continuous wavelength range is 0.4-2.5μm.

3. The material detection system based on photoacoustic effect according to claim 1, characterized in that, The dispersive component includes: a first grating, a second grating, and a slit stop; The first grating and the second grating are arranged opposite to and parallel to each other, and the first grating is configured to propagate the first detection light to the second grating; The second grating is configured to propagate the first detection light to the slit aperture; The slit stop is configured to filter out the second detection light from the first detection light; wherein the slit stop is configured to be movable relative to the second grating.

4. The material detection system based on photoacoustic effect according to claim 1, characterized in that, The focusing assembly includes a first galvanometer, a second galvanometer, and a field lens; The first galvanometer is disposed in the optical path of the second detection light, and the first galvanometer is configured to propagate the second detection light to the second galvanometer. The second galvanometer is configured to propagate the second detection light to the field mirror; The field lens is configured to focus the second detection light onto the surface of the material to be detected.

5. The material detection system based on photoacoustic effect according to claim 1, characterized in that, The acquisition component includes multiple collectors arranged in an array. The collector is configured to receive the acoustic signal and send the acoustic signal to the processor.

6. The material detection system based on photoacoustic effect according to claim 3, characterized in that, The length of the slit aperture is 0.15-1mm.

7. The material detection system based on photoacoustic effect according to claim 1, characterized in that, The focusing component also includes a driving element, which is connected to the first galvanometer and the second galvanometer respectively. The drive is configured to control the first and second galvanometers to swing in order to adjust the propagation direction of the second detection light.

8. The material detection system based on photoacoustic effect according to any one of claims 1-7, characterized in that, The processor includes an acquisition module, an establishment module, a training module, and a recognition module; The acquisition module is configured to acquire training data; The establishment module is configured to establish a recognition model; The training module is configured to train the recognition model using the training data; the input of the recognition model is the acoustic signal, and the output of the recognition model is the detection parameters of the material to be detected. The recognition module is configured to use the recognition model to identify the detection parameters of the material to be tested.

9. The material detection system based on photoacoustic effect according to claim 8, characterized in that, The training data includes a training set, a validation set, and a test set; The training set is configured to train the recognition model; The validation set is configured to validate the recognition model; The test set is configured to test the recognition model; wherein the ratio of the training set, the validation set, and the test set is 4:3:

3.

10. A material detection method based on photoacoustic effect, characterized in that, The material detection method, applied to the photoacoustic effect-based material detection system according to any one of claims 1-9, comprises: Send the first detection light; The first detection light is received and dispersed to obtain the second detection light; The second detection light is focused onto the surface of the material to be detected; Collect acoustic signals; the acoustic signals are generated by the second detection light acting on the material to be detected; The detection parameters of the material to be tested are determined based on the acoustic signal; the detection parameters include one or more of the following: material, surface process, impurities or damage.

Images