Automatic grading device for thermal shock cracks of coating and application method

Abstract

The invention discloses an automatic grading device for thermal shock cracks of a coating and an application method. Comprising a terahertz wave detection module, a laser ultrasonic excitation and detection module, a high-precision three-dimensional motion control platform, a multi-mode signal synchronous acquisition and processing unit, a crack intelligent grading evaluation unit, an environment simulation and temperature control unit, a protective shell and a human-computer interaction interface. The near-surface dielectric property change of the coating is represented through the terahertz technology, the deep mechanical property evolution is detected through the laser ultrasonic technology, the thermal shock service condition is reproduced in combination with a high-temperature environment simulation unit, and a multi-modal feature fusion algorithm and a physical model and machine learning collaborative inversion model are adopted to obtain the thermal shock resistance of the coating. And automatic quantitative evaluation and severity grading of the depth, density, length and network connectivity of the thermal shock cracks of the thermal barrier coating are realized. The problems that an existing single detection method is insensitive to vertical cracks, poor in high-temperature environment adaptability and low in manual evaluation efficiency are solved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of non-destructive testing of hot end components of an aero-engine, and particularly relates to a coating thermal shock crack automatic grading device based on terahertz and laser ultrasonic and application. BACKGROUND

[0002] A thermal barrier coating is a core protective layer of a turbine blade of an aero-engine, which is composed of a metal bonding layer (MCrAlY alloy) and a ceramic thermal insulation layer (8% yttria stabilized zirconia, 8YSZ). The thermal barrier coating is prone to generate a crack network perpendicular to the interface under high temperature, high temperature gradient and thermal cycle load due to thermal stress concentration. The depth, density and connectivity of the cracks directly determine the critical failure state of the coating. If the cracks extend to the blade substrate, the strength of the component will decrease sharply, and even the whole component will be scrapped, which will cause a major safety accident. Traditional destructive detection needs to cut off the blade sample for metallographic dissection, which is low in efficiency and cannot realize in-situ evaluation. Existing non-destructive detection technologies such as terahertz, ultrasonic and infrared thermal imaging have significant limitations. The terahertz technology is sensitive to near-surface parallel cracks but difficult to quantify deep vertical cracks (for example, the patent CN109490244A only evaluates the crack density by using the reflection peak amplitude ratio, the high-temperature signal-to-noise ratio is low, and the detection rate of vertical cracks is less than 30%). The laser ultrasonic technology can detect deep defects (for example, the patent CN113962124B optimizes laser parameters to avoid ablation), but is greatly affected by surface roughness, has poor curved surface adaptability and limited micron-level crack resolution. The infrared thermal imaging can only qualitatively evaluate the layered defects and cannot realize quantitative grading of cracks. Therefore, there is an urgent need for a detection device that can integrate multi-physical field information, adapt to high-temperature environment and realize automatic grading, so as to guarantee the service safety of hot end components of an aero-engine.

[0003] The existing patent technology has systematic defects in detection dimension, environmental adaptability and intelligence. In terms of curved surface adaptability, patent CN116067872A proposes a design of adjustable probe angle based on three-dimensional positioning, but relies on a single terahertz mode, with a vertical thermal shock crack detection rate of less than 30%, and does not integrate a temperature simulation module, which cannot reproduce the real morphology of cracks under thermal shock conditions. In terms of multi-modal fusion, patent CN116412767A develops a terahertz thickness measurement model based on the first three peaks (with an accuracy of microns), but it only targets a single parameter of coating thickness and does not involve crack feature extraction, and does not combine with ultrasonic mechanical property detection, resulting in insufficient accuracy of three-dimensional crack reconstruction; patent CN51675103 focuses on quantitative evaluation of TGO growth, but the technical route still mainly relies on a single mode, which is limited in detecting deep micro-cracks. In terms of intelligent grading, patent CN117871458A uses swarm intelligence algorithm to optimize terahertz feature extraction, but the model only targets debonding defects and does not cover multiple features such as depth, branch morphology and network connectivity of thermal shock cracks, and the device relies on manual adjustment of sample position, resulting in low scanning efficiency. In addition, the above-mentioned patents do not solve the problem of high-temperature in-situ detection, such as patent CN109490244A which only supports room temperature detection and cannot obtain the dynamic behavior of cracks under thermal expansion stress.

[0004] In view of the above technical bottlenecks, the present application proposes an integrated solution that combines multi-modal sensing, environmental simulation and intelligent evaluation. By capturing the changes in near-surface dielectric properties (sensitive to parallel cracks) using terahertz technology and characterizing deep mechanical properties (sensitive to vertical cracks) using laser ultrasonic technology, a high-precision three-dimensional motion platform is used to realize spatial registration of multi-modal data; an integrated resistance heating furnace and forced cooling system is used to simulate a thermal shock cycle environment of up to 1200°C, solving the problem of high-temperature in-situ signal synchronous acquisition; a crack inversion algorithm based on physical model (such as the terahertz wave layered medium propagation equation coupled with ultrasonic elastic wave scattering theory) is used to construct a quantitative mapping of crack parameters and multi-modal features, and a machine learning model is used to learn the grading rules of crack depth, density and network connectivity, finally outputting a three-dimensional crack distribution map and a quantitative grading report. This device will break through the limitations of existing patents such as single-modal detection bias, lack of high-temperature environment and reliance on human experience, and provide a high-precision, fully automated evaluation tool for coating failure warning and life prediction of aircraft engine blades, significantly reducing the risk of flight accidents and economic losses caused by coating peeling. SUMMARY

[0005] The present application aims to provide a coating thermal shock crack automatic grading device based on terahertz and laser ultrasound, which can realize the automatic quantitative evaluation and grading of the depth, density, length and network connectivity of the thermal barrier coating crack in a high-temperature environment simulating the actual thermal shock cycle through the multimodal data fusion of terahertz waves and laser ultrasound and the artificial intelligence algorithm. The device solves the problems of one-sidedness of single-mode detection, poor adaptability to high-temperature environment and strong dependence on artificial grading in the prior art, and provides a precise nondestructive evaluation means for the service safety of the coating of the turbine blade of an aero-engine.

[0006] To achieve the above-mentioned purpose, the present application adopts the following technical scheme: a coating thermal shock crack automatic grading device, comprising:

[0007] The terahertz wave detection module has a femtosecond laser with a wavelength of 1560 nm as the core, and the output pulsed laser pumps a photoconductive antenna to generate a wideband pulse of 0.1 to 3 THz after beam splitting, and the focused pulse is incident on the surface of the sample in the form of a 2 mm diameter spot through a parabolic mirror;

[0008] The reflection signal of the detection module is collected by another set of parabolic mirrors and introduced into a zinc telluride crystal detector, and the time-domain waveform is scanned by an optical delay line with a step of 0.1 fs to obtain the dielectric constant distribution information of the coating, and the sensitivity to parallel cracks within 50 microns near the surface reaches 10 microns.

[0009] The laser ultrasound excitation and detection module adopts a picosecond pulsed laser with a wavelength of 1064 nm, and the laser beam with an output energy of 50 muJ is shaped into a 1*5 mm line light source by a cylindrical lens to irradiate the sample, and a wideband ultrasonic wave with a center frequency of 15 MHz is excited through the thermoelastic effect;

[0010] The sound wave propagation signal is received by a Doppler laser interferometer to detect the surface vibration with a displacement resolution of 50 nm, and the 1 to 30 MHz frequency band characteristics are extracted by combining a band-pass filter, and the positioning accuracy of the vertical cracks within 200 microns in the deep layer reaches 20 microns.

[0011] The high-precision three-dimensional motion control platform is driven by a servo motor and carries an optical encoder with a resolution of 0.1 microns, and has a positioning accuracy of 10 microns and a repeat positioning accuracy of 5 microns;

[0012] Preferably, the control platform moves the sample at a speed of 20 mm / s along the preset scanning path, ensures that the terahertz focal point and the laser ultrasound action point have a deviation of less than 5 microns in the three-dimensional space, and realizes the pixel-level registration of the multimodal data.

[0013] The multimodal signal synchronous acquisition and processing unit synchronously captures the terahertz time-domain electric field signal and the ultrasonic time-domain vibration signal by using a high-speed acquisition card with a sampling rate of 10 GSPS;

[0014] The digital signal processor runs a wavelet denoising algorithm, adopts db8 wavelet basis to perform 5-layer decomposition to eliminate thermal noise, and extracts an ultrasonic dispersion curve through a short-time Fourier transform;

[0015] Preferably, the feature extraction module calculates the phase delay in the terahertz signal with an accuracy of 0.1 radian, and the pulse broadening rate with a resolution of 0.1 ps, while extracting the sound attenuation coefficient in the ultrasonic signal with a dynamic range of 60 dB and a 2 MHz characteristic frequency amplitude value.

[0016] The environmental simulation and temperature control unit integrates a silicon carbide resistance heating furnace to simulate a thermal shock process by heating at a rate of 20 ℃ / s to 1200 ℃ constant temperature;

[0017] The liquid cooling system is connected to cooling liquid below -10 ℃ to achieve rapid cooling at 100 ℃ / s, and the cycle period can be set to a combination of heating for 300 s and cooling for 60 s.

[0018] The platinum-rhodium thermocouple feeds back the surface temperature in real time, and the closed-loop control accuracy is ±5 ℃.

[0019] The crack intelligent grading evaluation unit first fuses the terahertz phase delay feature and the ultrasonic sound attenuation coefficient, inputs a physics-based inversion model, and the model establishes a coupled model of Maxwell equations and elastic wave equations to solve the crack depth and width.

[0020] The feature set is imported into a pre-trained random forest model, the training sample size is 10,000 groups, the input feature dimension is 12, the output crack depth grade is defined as grade I 0-50 μm, grade II 50-100 μm, grade III greater than 100 μm, the density grade is defined as low density less than 5 / mm 2 , medium density 5-10 / mm 2 , high density greater than 10 / mm 2 , and network connectivity index 0-1 continuous value;

[0021] Finally, a three-dimensional crack distribution map and a quantitative grading report are generated.

[0022] The protective shell of the automatic grading evaluation device adopts 1 mm thick permalloy to achieve 60 dB electromagnetic shielding, and the observation window is coated with a 1064 nm and 1560 nm dual-band laser protection film with an optical density OD greater than 7.

[0023] The human-computer interaction interface is a 10-inch touch screen that supports setting the scanning area with a minimum step of 5 μm, the thermal shock program can customize the temperature curve, and the model parameters support adjustable physics inversion weight 0.3-0.7; real-time display of terahertz time-domain waveform, ultrasonic spectrum and temperature curve; dynamically rendering the three-dimensional reconstruction result of the crack and labeling the grade with pseudo-color.

[0024] The application method of the device comprises the following steps:

[0025] The turbine blade sample after thermal shock test is fixed on the motion platform, the scanning area is set as 20*20mm through the touch screen 2 , the scanning step is 50um; the heating furnace is started to preheat the sample to 300℃ at a rate of 15℃ / s, and the thermal stress interference is eliminated;

[0026] Then, the thermal shock cycle is performed, the temperature is raised to 1200℃ at a rate of 15℃ / s and kept for 180s, and the liquid cooling system is started and reduced to 200℃ within 25s.

[0027] The motion platform moves the first measuring point of the sample to the confocal point position, synchronously triggers the terahertz module, the scanning range is 15ps, the resolution is 0.2ps, and the laser ultrasonic module, the excitation energy is 30mJ / cm 2 ;

[0028] The acquisition unit synchronously captures the terahertz time-domain waveform and ultrasonic vibration signal at a rate of 100 times per second.

[0029] The signal processing unit performs wavelet denoising on the original data, the threshold is set as 0.05, and feature extraction: the terahertz pulse broadening rate is calculated, the reference value is 1.5ps, and the measured value is 2.8ps, and the ultrasonic 2MHz amplitude attenuation amount, the reference value is -20dB, and the measured value is -35dB.

[0030] The motion platform moves to the next measuring point according to the preset path, and the acquisition is repeated until the full area scanning is completed.

[0031] The hierarchical evaluation unit fuses all the features of the measuring points, the physical model calculates the crack depth, and the random forest model judges the depth, high density and connectivity index grade;

[0032] A thermal map containing crack spatial distribution and a statistical report are generated, including maximum depth and density information, and a moderate damage conclusion is output through the touch screen.

[0033] Beneficial effects

[0034] Compared with the prior art, the advantages of the present application are:

[0035] 1. The present application adopts multi-modal collaborative detection, through the spatial registration of terahertz and laser ultrasonic, the error is less than 5um, the dielectric properties and mechanical property parameters are synchronously acquired, the defect that the existing patent is not sensitive to vertical cracks is overcome, and the crack detection rate is improved to 95%.

[0036] 2. The application realizes high-temperature in-situ evaluation, integrates a 1200°C environment simulation module, and the existing patents do not involve this function. The liquid cooling system realizes a 100°C / s quenching rate, accurately reproduces thermal shock stress conditions, and for the first time realizes in-situ capture of the dynamic behavior of cracks under thermal expansion.

[0037] 3. The application constructs an intelligent hierarchical system, combines physical models and machine learning, and the existing patents generally use a single algorithm. A multi-dimensional hierarchical standard of depth-density-connectivity is established, and a quantitative report is output to replace manual experience judgment, and the evaluation efficiency is improved by 10 times.

[0038] 4. The application emphasizes the deep integration of terahertz-ultrasonic multi-physical field coupling detection, thermal shock environment simulation and intelligent grading algorithm, provides a high-precision and high-efficiency solution for thermal barrier coating life evaluation, and significantly reduces the risk of safety accidents caused by coating failure of an aero-engine. BRIEF DESCRIPTION OF DRAWINGS

[0039] The accompanying drawings, which form a part of this application, are included to provide a further understanding of the application and are incorporated in and constitute a part of this application. The illustrative embodiments of the application, and their description, serve to explain the application. The drawings should not be used to limit the application unduly.

[0040] Figure 1 It is a whole structure schematic diagram of the multi-modal evaluation device of the embodiment of the application.

[0041] Figure 2 It is a high-precision three-dimensional motion control platform and environment simulation unit structure schematic diagram of the embodiment of the application.

[0042] Figure 3 It is a man-machine interface layout schematic diagram of the embodiment of the application.

[0043] Figure 4 It is a thermal shock crack multi-modal detection and intelligent grading evaluation flowchart of the embodiment of the application.

[0044] 1, protective cover; 2, THz system; 3, heating system; 4, laser ultrasonic system; 5, thermal barrier coating sample; 6, man-machine interface; 7, laser protection observation window; 8, cooling system; 9, high-precision three-dimensional motion platform; 10, heat preservation shell; 11, clamp; 12, sample; 13, three-dimensional motion platform X axis; 14, three-dimensional motion platform Z axis; 15, three-dimensional motion platform Y axis; 16, horizontal base; 17, cooling liquid pipeline; 18, telescopic cooling gas pipe; 19, temperature sensor; 20, temperature sensor control line; 21, temperature sensor lifting frame. DETAILED DESCRIPTION

[0045] It needs to be clear that the embodiments and their technical features proposed in the present application can form new technical solutions in a complementary combination manner without contradiction. The following will be based on the technical architecture shown in the drawings, combined with specific embodiments to develop systematic description, to present the synergistic effect and implementation details in different scenarios.

[0046] It needs to be pointed out that the steps marked in the flowchart of all the drawings of the present application are essentially operation logic constructed based on computer executable instruction set. Although the steps in the flowchart are arranged and displayed in a certain topological order, in the actual computer system execution process, according to the specific running environment, resource scheduling state and exception handling mechanism, there is a technical possibility of dynamically adjusting the execution order.

[0047] The present application provides a kind of coating thermal shock crack automatic grading device, as shown in Figure 1 It mainly includes protective cover 1, THz system 2, heating system 3, laser ultrasonic system 4, thermal barrier coating sample 5, man-machine interface 6, laser protection observation window 7, cooling system 8 and high-precision three-dimensional motion platform 9 and other core components.

[0048] The device is integrated in a protective cover 1 made of permalloy, which provides 60dB electromagnetic shielding capability to ensure that the terahertz signal is not disturbed by the outside world. The front of the device is provided with a laser protection observation window 7, which uses a 1064nm and 1560nm dual-band protection filter to ensure safe operation. The man-machine interface 6 is located on the right side of the front of the device, which is designed with a 10-inch touch screen, which is used to control the entire detection process and can visualize the results.

[0049] The operation of the device is based on the principle of multi-modal fusion detection.

[0050] First, the thermal barrier coating sample 5 is installed on the high-precision three-dimensional motion platform 9 and fixed by the clamp 11. The operator sets the detection parameters through the man-machine interface 6, including the scanning area range, the temperature curve of thermal shock cycle 20℃ / s heating to 1200℃, 300s holding, 100℃ / s cooling and other parameters.

[0051] After starting the detection program of the device, the heating system 3 and the cooling system 8 start to work to simulate the actual thermal shock environment.

[0052] The THz system 2 and the laser ultrasonic system 4 are synchronously scanned and detected under the unified clock.

[0053] The multi-modal signal acquisition and processing unit processes the collected signals in real time, and the crack intelligent grading evaluation unit analyzes the fused features to finally generate a crack evaluation report.

[0054] The high-precision three-dimensional motion platform 9 is driven by a servo motor and equipped with an optical encoder with a resolution of 0.1 μm, ensuring a positioning accuracy of 10 μm.

[0055] The platform adopts an orthogonal structure of X-axis 13, Y-axis 15 and Z-axis 14 and is installed on a horizontal base 16.

[0056] The sample clamp 11 is made of high-temperature-resistant ceramic material and can stably fix turbine blade samples of various curved shapes.

[0057] The cooling system 8 comprises a cooling liquid pipeline 17 and a retractable cooling gas pipeline 18 and adopts a liquid-gas mixed cooling mode to achieve rapid cooling.

[0058] The temperature sensor 19 is connected to the main controller through a temperature sensor control line 20, monitors the sample temperature in real time, and adjusts the measurement position through a temperature sensor lifting frame 21.

[0059] The THz system 2 adopts a 1560 nm femtosecond laser as a light source to generate a broadband pulse of 0.1 to 3 THz.

[0060] The THz system comprises a terahertz transmitter, a receiver and an optical delay line, and focuses to form a 2 mm diameter spot through a parabolic mirror.

[0061] The laser ultrasonic system 4 adopts a 1064 nm picosecond pulse laser with an output energy of 50 μJ and shapes it into a 1×5 mm linear light source through a cylindrical mirror.

[0062] The receiving part of the laser system adopts a Doppler laser interferometer with a displacement resolution of 50 nm. The light paths of the two systems are precisely calibrated to ensure that the focal point deviation on the sample surface is less than 5 μm.

[0063] The heating system 3 adopts a silicon carbide resistance heating furnace with a maximum temperature of 1200℃. The furnace body is coated with a heat preservation shell 10 to reduce heat loss.

[0064] The cooling system 8 delivers 10℃ below zero cooling liquid through the cooling liquid pipeline 17 and realizes a rapid cooling rate of 100℃ / s in cooperation with the airflow cooling of the retractable cooling gas pipeline 18.

[0065] The temperature sensor 19 adopts a platinum-rhodium thermocouple with a measurement accuracy of ±5℃.

[0066] The multi-modal signal acquisition and processing unit adopts a high-speed acquisition card with a sampling rate of 10 GSPS to synchronously acquire the terahertz time-domain signal and the ultrasonic vibration signal.

[0067] Further signal processing adopts db8 wavelet base to carry out 5-layer decomposition and noise reduction, extracts the phase delay and pulse broadening characteristics of terahertz signals, and the sound attenuation and 2MHz frequency amplitude characteristics of ultrasonic signals.

[0068] Further, the crack intelligent grading evaluation unit adopts a combination of a physical inversion model and a machine learning model. The physical model is based on Maxwell's equations and elastic wave theory, and the machine learning model uses a random forest algorithm with a training sample size of 10,000 groups.

[0069] The overall operation process of the device is as follows:

[0070] S1, install the thermal barrier coating sample to the high-precision three-dimensional motion platform and clamp it. The operator needs to ensure that the sample is firmly fixed by the clamp 11, and the surface of the sample is parallel to the motion platform.

[0071] S2, set the detection parameters through the human-computer interaction interface 6, including setting a 20mm x 20mm scanning area, a 50μm scanning step, and thermal shock cycle program parameters including a 20℃ / s temperature rise to 1200℃, a 300s holding period, and a 100℃ / s cooling to 200℃.

[0072] S3, start the environmental simulation and temperature control unit according to the preset thermal shock cycle program. The heating system 3 starts to work at a rate of 20℃ / s, and the temperature sensor 19 monitors the sample temperature in real time. After reaching 1200℃, the system enters the holding stage, and after 300s, the cooling system 8 is started to work at a rate of 100℃ / s.

[0073] S4, the high-precision three-dimensional motion platform 9 moves the sample to be tested to the common focal point position of the THz system 2 and the laser ultrasonic system 4. The platform moves at a speed of 20mm per second according to the preset scanning path, ensuring that the positioning accuracy of each detection point reaches 10μm.

[0074] S5, trigger the terahertz and laser ultrasonic modules synchronously, and the high-speed acquisition card synchronously acquires the original time domain signal. The terahertz system acquires a time domain waveform in the range of 15ps with a resolution of 0.2ps; the laser ultrasonic system acquires a vibration signal with a bandwidth of 30MHz and a sampling rate of 100 times per second.

[0075] S6, signal processing is carried out using wavelet transform, and db8 wavelet base is used for 5-layer decomposition with a threshold value of 0.05 to effectively eliminate thermal noise and environmental noise. The signal-to-noise ratio of the processed signal is improved by more than 20dB.

[0076] S7, feature extraction is performed on the processed signal. The phase delay feature is extracted from the terahertz signal with an accuracy of 0.1 radian, and the pulse broadening feature is extracted with a resolution of 0.1 ps. The sound attenuation coefficient is extracted from the ultrasonic signal with a dynamic range of 60 dB, and the amplitude value of the 2 MHz characteristic frequency is extracted.

[0077] S8, determine whether the entire area has been scanned. If not, return to step S4 to continue scanning; if yes, proceed to the next step. Typically, a 20mm x 20mm area scan takes about 30 minutes.

[0078] S9, multimodal feature fusion of terahertz and ultrasonic features. A feature layer fusion method is used to consider the sensitivity differences of different modalities to crack features, generating a 12-dimensional fusion feature vector.

[0079] S10, construct a physical inversion model based on Maxwell's equation and elastic wave equation, and a machine learning model based on random forest algorithm. The physical model solves the quantitative relationship between crack depth and width, and the machine learning model processes complex features such as crack density and network connectivity.

[0080] S11, decision fusion of physical inversion model and machine learning model to calculate crack parameters. Weighted average method is used for decision fusion, physical model weight 0.3, machine learning model weight 0.7. The output crack depth classification results include I class 0-50 μm, II class 50-100 μm, III class 100 μm above; density classification includes low density less than 5 per square millimeter, medium density 5-10 per square millimeter, high density greater than 10 per square millimeter; network connectivity index 0-1 continuous value.

[0081] S12, generate a comprehensive evaluation report containing crack distribution map, quantitative parameters and classification conclusions, etc. The report includes two-dimensional and three-dimensional crack distribution map, maximum crack depth, average crack density, network connectivity index and other parameters, as well as comprehensive classification conclusions based on these parameters.

[0082] S13, visual output of crack detection results through human-computer interaction interface 6. Touch screen displays real-time signal waveform, processed feature curve, temperature change curve, and final three-dimensional crack distribution map and classification report. The operator can save the detection data, export the detection report, or adjust the parameters for re-detection through the interface.

[0083] The core innovation of the device is to realize the multimodal fusion detection of terahertz and laser ultrasound. Combined with high temperature environment simulation and intelligent grading algorithm, it can comprehensively and quantitatively evaluate the thermal barrier coating thermal shock crack. Compared with single detection method, the crack detection rate is improved to 95%, and the evaluation efficiency is improved by 10 times, providing a reliable technical means for safety evaluation of aero-engine hot end components.

Claims

1. A coated thermal shock crack automatic grading apparatus characterized by, The application relates to a thermal barrier coating crack detection and evaluation system. The system comprises a terahertz wave detection module, a laser ultrasonic excitation and detection module, a high-precision three-dimensional motion control platform, a multi-modal signal synchronous acquisition and processing unit, a crack intelligent grading evaluation unit, an environment simulation and temperature control unit, a protective shell and a man-machine interface. The terahertz wave detection module is used for emitting wide-band terahertz pulses to the surface and interior of a thermal barrier coating sample to be detected and receiving reflected or transmitted signals, and the near-surface and subsurface cracks are sensed by the interaction between terahertz waves and the spatial distribution change of the dielectric constant of the material. The laser ultrasonic excitation and detection module is used for inducing wide-band ultrasonic waves on the surface of the sample by pulsed laser, and the propagation and scattering signals of the ultrasonic waves in the coating and the substrate are detected by a non-contact optical receiving device, and the reflection, refraction and scattering characteristics of the ultrasonic waves at the material acoustic impedance mutation interface are used to characterize the deep crack morphology and distribution. The high-precision three-dimensional motion control platform carries the sample and realizes the precise displacement of the sample in three orthogonal directions in space, so that the terahertz wave detection point and the laser ultrasonic action point are accurately overlapped or scanned according to a preset path. The multi-modal signal synchronous acquisition and processing unit is responsible for high-speed synchronous acquisition, noise reduction, enhancement and feature extraction of the terahertz time-domain waveform signals and the laser ultrasonic time-domain and frequency-domain signals. The crack intelligent grading evaluation unit realizes the automatic quantitative evaluation and severity grading of crack depth, length, density and network connectivity based on the fused terahertz feature parameters and ultrasonic feature parameters, the thermal shock crack physical model and the machine learning algorithm. The environment simulation and temperature control unit is used for simulating thermal shock cycles or specific temperature environments during the evaluation process; the protective shell provides electromagnetic shielding and laser safety protection; and the man-machine interface is used for parameter setting, process monitoring and evaluation result visual output.

2. The apparatus of claim 1, wherein, The terahertz wave detection module specifically comprises a femtosecond laser, a terahertz emitter, a terahertz receiver, a terahertz optical delay line, a focusing parabolic mirror and a terahertz detector. The pulsed laser output by the femtosecond laser is split and used to pump the terahertz emitter to generate wide-band terahertz pulses and as the gating light of the terahertz detector; the terahertz pulses are collimated and focused by the parabolic mirror and then incident on the sample surface, and the reflected or transmitted terahertz waves carrying the internal structure information of the sample are collected by another set of parabolic mirrors and then focused on the terahertz detector. The terahertz optical delay line realizes point-by-point scanning measurement of the terahertz time-domain waveform by precisely controlling the optical path difference between the pumping light and the detection light, and obtains the time-domain electric field intensity change information of the sample in the terahertz frequency band, which is directly related to the dielectric constant distribution of the sample and is highly sensitive to the dielectric mutation height of the air-material interface caused by thermal shock cracks.

3. The apparatus of claim 1, wherein, The laser ultrasonic excitation and detection module specifically comprises a high-energy nanosecond or picosecond pulsed laser, an optical system for laser beam shaping and focusing and a non-contact laser ultrasonic receiving device. The high-energy laser beam output by the pulsed laser ultrasonic emitter is shaped and focused into a specific spot shape by the optical system, and then irradiated on the surface of the sample, and wide-band ultrasonic waves are instantaneously excited in the irradiation area through the thermoelastic effect or the ablation effect. The wideband ultrasonic wave propagates into the sample, and when encountering a crack defect, it is reflected, scattered and mode-converted; The non-contact laser ultrasonic receiving device uses a laser interferometer or a beam deflection method to detect the transient out-of-plane or in-plane displacement vibration of the sample surface caused by the arrival of the ultrasonic wave, converts the acoustic signal into an optical signal for high-sensitivity reception, and obtains the characteristic information of the ultrasonic wave in the time domain and the frequency domain, which deeply reflects the changes of the material elastic modulus, density and the hindering effect of the crack on the propagation of the acoustic wave.

4. The apparatus of claim 1, wherein, The high-precision three-dimensional motion control platform is driven by a precision stepping motor or a servo motor and is equipped with a high-resolution optical encoder for position feedback, with a positioning accuracy better than 10 μm and a repeat positioning accuracy better than 5 μm; the platform programming control realizes point measurement, line scanning or surface scanning mode, and the platform design ensures that the terahertz beam focal point, the laser ultrasonic excitation point and the detection point are consistent in space at a specified depth on the sample surface or inside.

5. The apparatus of claim 1, wherein, The multi-modal signal synchronous acquisition and processing unit includes a high-speed data acquisition card, a digital signal processor and a signal processing algorithm; the high-speed data acquisition card synchronously acquires the time-domain electric field waveform signal from the terahertz detector and the time-domain acoustic vibration signal from the laser ultrasonic receiving device under the trigger of a unified clock source; The digital signal processor runs a denoising algorithm, a time-frequency analysis algorithm and a feature extraction algorithm; The denoising algorithm eliminates environmental noise and system noise based on wavelet transform or empirical mode decomposition; The time-frequency analysis algorithm obtains the frequency dispersion characteristics of the ultrasonic signal through short-time Fourier transform or wavelet transform; The feature extraction algorithm extracts time-domain peak value, pulse broadening, spectral amplitude, phase delay and other features from the terahertz signal, and extracts sound speed, sound attenuation coefficient, specific frequency component amplitude, time difference of arrival and echo feature physical parameters from the ultrasonic signal.

6. The apparatus of claims 1 and 5, wherein, The core of the crack intelligent grading evaluation unit is a multi-modal feature fusion and physically modeled intelligent evaluation algorithm; The algorithm first fuses the extracted terahertz feature parameter set and the laser ultrasonic feature parameter set at the feature layer or the decision layer, and the fusion process considers the sensitivity differences of different modalities to different dimensional properties of the crack; Subsequently, the fused features are input into a physically based crack inversion model or a pre-trained machine learning model, which uses the propagation equation of terahertz waves in layered media combined with the crack scattering model, and the elastic wave propagation theory in the medium containing defects to construct the quantitative mapping relationship between the crack parameters and the multi-modal signal features; The machine learning model is trained using a large number of sample data with known crack grades to learn the complex nonlinear mapping relationship from the fused features to the crack depth grade, length grade, density grade and network connectivity grade; Finally, an evaluation report containing a crack two-dimensional or three-dimensional distribution map and a quantitative grading result is output.

7. The apparatus of claim 1, wherein The environment simulation and temperature control unit comprises a heating furnace, a cooling system, a temperature sensor and a controller. The heating furnace adopts resistance heating or infrared radiation heating method to apply programmed temperature heating cycle or constant high temperature environment to the sample, and the maximum temperature is not less than 1200℃. The cooling system adopts forced air cooling or liquid cooling method to realize rapid cooling and simulate thermal shock process. The temperature sensor monitors the surface or near-surface temperature of the sample in real time and feeds back to the controller to realize accurate closed-loop control of temperature. The control unit enables the device to evaluate the coating cracks in situ under the simulated actual service thermal shock conditions or at a specific temperature point.

8. The apparatus of claim 1, wherein, The protective shell is composed of an electromagnetic shielding material and a laser protection observation window. The electromagnetic shielding material effectively isolates the disturbance of external environment electromagnetic interference on the weak terahertz signal, and prevents the leakage of electromagnetic radiation generated inside the device. The laser protection observation window uses a specific wavelength protection filter to ensure the safe observation of the sample state and the laser light path by the operator.

9. The apparatus of claim 1, wherein, The human-computer interaction interface is an integrated touch screen or connected external computer software, which provides the following functions: Setting terahertz scanning parameters, laser ultrasonic excitation energy and receiving parameters, motion platform scanning path and speed, temperature control program, signal processing parameters, evaluation model selection; Real-time display of original signal waveform, processed signal characteristics, platform position information and temperature curve; Dynamic visualization of crack images after multi-modal data fusion and final grading evaluation results, including crack distribution map, size statistics, depth profile and severity grade label.

10. The method of using the device of claim 1, wherein, The following automatic steps are included: (1) Fixing the thermal barrier coating sample after thermal shock test on the three-dimensional motion platform; (2) Setting evaluation parameters and scanning area through the human-computer interaction interface; (3) Starting the environment simulation unit to make the sample reach the preset temperature; (4) The motion platform moves the sample to be tested to the terahertz and laser ultrasonic focal point position; (5) Synchronously triggering the terahertz wave detection module and the laser ultrasonic excitation module to collect the terahertz time domain signal and the laser ultrasonic time domain signal of the sample respectively; (6) The signal acquisition and processing unit synchronously collects, denoises, enhances and extracts multi-modal physical feature parameters from the original signal; (7) The motion platform moves to the next measurement point according to the preset path to repeat the collection until the entire area scanning is completed; (8) The crack intelligent grading evaluation unit inputs the fused multi-modal features of each point into the evaluation model to calculate the crack feature parameters of each spatial position and grade the severity accordingly; (9) Finally, a comprehensive evaluation report containing crack spatial distribution map, quantitative parameter statistics and automatic grading results is generated and output through the human-computer interaction interface.

Images