Laser ultrasonic detection method and system for metal grain size under high-temperature dynamic working condition

Abstract

The invention provides a high-temperature dynamic working condition metal grain size laser ultrasonic detection method and system, and relates to the field of data processing. By constructing a programmable high-temperature environment and introducing controllable reciprocating motion, in-situ laser ultrasonic detection of multiple samples under a high-temperature dynamic working condition is realized; the two-dimensional mobile platform is used for ensuring stable alignment of the excitation light spot and the detection light spot in the movement process, and ultrasonic signals, temperature working condition information and speed working condition information are synchronously collected; time domain and frequency domain attenuation features are extracted based on a multi-working-condition standard sample, a high-dimensional feature vector is constructed, and an artificial intelligence model between ultrasonic attenuation and the grain size is trained; in actual detection, a grain size prediction result and a confidence interval are output, the grain size grade is converted, and batch in-situ lossless quantitative detection of the grain size under the high-temperature dynamic condition is achieved. By means of the technical scheme, non-contact, in-situ, continuous and quantitative batch grain size nondestructive testing can be conveniently achieved under the high-temperature dynamic working condition.

Description

Technical Field

[0001] This application relates to the field of data processing technology, specifically to a laser ultrasonic detection method and system for metal grain size under high-temperature dynamic conditions. Background Technology

[0002] During the high-temperature manufacturing and service processes of metallic materials, dynamic factors such as motion loads, stress coupling, or rapid changes in operating conditions often occur simultaneously. The internal grain size and microstructure of these materials will continue to evolve over time. As a key microscopic parameter affecting the strength, plasticity, fatigue life, and service reliability of materials, accurate monitoring of grain size is of great engineering significance.

[0003] Current grain size detection methods mainly rely on metallography and electron backscatter diffraction (ESD). These methods generally require destructive pretreatment of samples, such as sampling, cutting, and polishing. Furthermore, the detection process can only be performed offline under normal temperature and static conditions, resulting in low efficiency and long cycles, making it difficult to meet the needs of real-time grain size assessment under high-temperature dynamic conditions. Simultaneously, during actual material service or processing, grain evolution often exhibits significant time-varying and spatial non-uniformity. Traditional offline detection methods cannot reflect the continuous evolution of grain size in high-temperature dynamic environments, nor can they provide in-situ characterization of grain states under motion or stress-coupled conditions, leading to significant deviations between the detection results and the actual service conditions. Therefore, none of the above methods can achieve non-contact, in-situ, continuous, and quantitative batch non-destructive grain size detection under high-temperature dynamic conditions.

[0004] Therefore, there is an urgent need for a laser ultrasonic testing method and system for metal grain size under high-temperature dynamic conditions. Summary of the Invention

[0005] This application provides a laser ultrasonic testing method and system for metal grain size under high-temperature dynamic conditions, which facilitates non-contact, in-situ, continuous, and quantitative batch non-destructive testing of grain size under high-temperature dynamic conditions.

[0006] The first aspect of this application provides a high-temperature dynamic condition laser ultrasonic testing method for metal grain size. The method includes: constructing a programmable high-temperature environment using a heating furnace with a transparent quartz window, a temperature controller, and an infrared thermometer; controlling the heating and holding of at least one metal sample clamped within the heating furnace according to the programmable high-temperature environment; and acquiring temperature information corresponding to a preset detection area on the surface of the metal sample; controlling a linear motor to drive the heating furnace and multiple metal samples inside to reciprocate at a uniform speed along a preset motion direction to form a controllable and repeatable dynamic condition; and simultaneously acquiring the temperature information of the metal sample during the detection process. The motion position and speed during the process are used as speed condition information; the two-dimensional moving platform is controlled to coordinately adjust the positions of the pulsed laser excitation head and the laser interferometer receiving head, so that the excitation spot and the detection spot always cover the preset scanning area during the high-temperature dynamic movement of the metal sample, thereby forming multiple stable detection points in the same detection process and realizing in-situ scanning detection of multiple samples; after the temperature condition information and the speed condition information reach a stable state, the synchronization control card is controlled to simultaneously trigger the nanosecond pulsed laser to excite broadband ultrasonic waves on the surface of the metal sample, and trigger the laser interferometer to receive the ultrasonic waves propagated in a non-contact manner. The surface micro-vibration echo signal is used to simultaneously control a high-speed data acquisition card to synchronously acquire and transmit the ultrasonic time series, temperature condition information, and speed condition information to an industrial control computer for storage and processing, thereby obtaining a laser ultrasonic detection signal containing multiple echoes and corresponding condition information. Using a pre-prepared standard sample of the same grade as the metal material to be tested but with different grain sizes and uniform grain distribution within the C-scan detection area, the in-situ laser ultrasonic detection process is reproduced under conditions covering room temperature to high temperature and various motion speeds. This allows for the extraction of time-domain and frequency-domain attenuation features from the laser ultrasonic detection signal. Based on the time-domain attenuation... The system constructs a high-dimensional feature vector by combining the characteristics, frequency domain attenuation features, and corresponding working condition information. This vector is then combined with a standard sample grain size monitoring label obtained in a preset manner to train an artificial intelligence model relating ultrasonic attenuation and grain size under high-temperature dynamic conditions. Laser ultrasonic acquisition and feature extraction are performed on the test sample under target high temperature and target motion speed conditions. The obtained target high-dimensional feature vector is input into the artificial intelligence model, which outputs the grain size prediction results and confidence intervals for each detection point and converts them into grain size grades. This enables batch in-situ non-destructive quantitative detection of grain size in metallic materials under high-temperature dynamic conditions.

[0007] A second aspect of this application provides a high-temperature dynamic condition laser ultrasonic testing system for metal grain size. The system includes an acquisition module and a processing module. The acquisition module is used to construct a programmable high-temperature environment using a heating furnace with a transparent quartz window, a temperature controller, and an infrared thermometer. Based on this programmable high-temperature environment, it performs heating and holding control on at least one metal sample clamped within the heating furnace, and acquires temperature condition information corresponding to a preset detection area on the surface of the metal sample. The acquisition module is also used to control a linear motor to drive the heating furnace and multiple metal samples inside it to reciprocate at a uniform speed along a preset motion direction, thereby forming a controllable and repeatable dynamic condition. The system synchronously acquires the movement position and speed of the metal sample during the detection process as speed condition information. The processing module controls the two-dimensional moving platform to coordinate the positions of the pulsed laser excitation head and the laser interferometer receiving head, ensuring that the excitation and detection spots always cover the preset scanning area during the high-temperature dynamic movement of the metal sample. This allows for the formation of multiple stable detection points and in-situ scanning detection of multiple samples during the same detection process. The processing module also controls the synchronization control card to simultaneously trigger a nanosecond pulsed laser to excite broadband ultrasonic waves on the surface of the metal sample after the temperature and speed condition information have reached a stable state, and triggers the laser interferometer to... The system receives surface micro-vibration echo signals caused by ultrasonic propagation in a non-contact manner. Simultaneously, a high-speed data acquisition card is controlled to synchronously acquire and transmit the ultrasonic time series, temperature condition information, and velocity condition information to an industrial control computer for storage and processing. This yields a laser ultrasonic detection signal containing multiple echoes and corresponding condition information. The processing module is further used to reproduce the in-situ laser ultrasonic detection process using pre-prepared standard samples of the same grade as the metal material to be tested but with different grain sizes and uniform grain distribution within the C-scan detection area, covering a range from room temperature to high temperature and various motion speeds. This allows for the extraction of time-domain and frequency-domain attenuation features from the laser ultrasonic detection signal. Based on the time-domain attenuation characteristics, the frequency-domain attenuation characteristics, and the corresponding operating conditions, a high-dimensional feature vector is constructed. Combined with the standard sample grain size monitoring label obtained in a preset manner, an artificial intelligence model is trained to form an ultrasonic attenuation and grain size relationship under high-temperature dynamic conditions. The processing module is also used to perform laser ultrasonic acquisition and feature extraction on the test sample under target high temperature and target motion speed conditions. The obtained target high-dimensional feature vector is input into the artificial intelligence model, and the grain size prediction results and confidence intervals corresponding to each detection point are output. The grain size level is then calculated to realize batch in-situ non-destructive quantitative detection of grain size of metallic materials under high-temperature dynamic conditions.

[0008] A third aspect of this application provides an electronic device including a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, and both the user interface and the network interface are used to communicate with other devices. The processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method described above.

[0009] A fourth aspect of this application provides a non-transitory computer-readable storage medium storing instructions that, when executed, perform the method described above.

[0010] In summary, one or more technical solutions provided in this application have at least the following technical effects or advantages: By introducing a programmable high-temperature environment and simultaneously acquiring temperature and speed information, each laser ultrasonic test has a clear and traceable condition label, thus establishing an explicit correlation between grain size changes and key operating conditions such as temperature and motion. This not only enhances the physical interpretability of the test results but also provides a reliable data foundation for subsequent analysis of grain evolution mechanisms and process optimization. Simultaneously, the collaborative design of the linear motor and the two-dimensional moving platform enables batch scanning tests of multiple samples and multiple test points under the same dynamic conditions, significantly improving testing efficiency and operating condition consistency, and avoiding the representativeness issues caused by single-point or single-sample testing. The non-contact laser ultrasonic excitation and reception method effectively avoids the problems of easy failure and unstable coupling of traditional contact sensors under high-temperature environments, ensuring good stability and repeatability of the detection system under high-temperature and motion conditions. Furthermore, the synchronous control card achieves strict time alignment between ultrasonic excitation, reception, and multi-source operating condition data, guaranteeing a one-to-one correspondence between ultrasonic detection signals and operating condition information such as temperature and speed, providing highly consistent raw data for subsequent feature extraction and modeling.

[0011] By training a standard sample system and an artificial intelligence model, the ultrasonic attenuation behavior under complex high-temperature dynamic conditions is transformed into a quantifiable high-dimensional feature expression. This allows grain size inversion to move beyond relying on single empirical formulas or static assumptions, enabling adaptive modeling that comprehensively utilizes time-domain, frequency-domain, and operating condition information. This significantly improves the accuracy and robustness of grain size prediction. Simultaneously, by outputting the grain size prediction results and their confidence intervals, and further converting them into commonly used engineering grain size grades, the detection results possess both quantitative accuracy and ease of engineering application and quality assessment. Overall, this achieves a comprehensive improvement in the accuracy, real-time performance, engineering usability, and batch testing capabilities of grain size detection under high-temperature dynamic conditions. Therefore, it facilitates non-contact, in-situ, continuous, and quantitative batch non-destructive testing of grain size under high-temperature dynamic conditions. Attached Figure Description

[0012] Figure 1 A schematic flowchart of a laser ultrasonic testing method for metal grain size under high-temperature dynamic conditions provided in this application embodiment; Figure 2 A schematic diagram of a high-temperature dynamic metal grain size laser ultrasonic testing system provided in this application embodiment; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0013] Explanation of reference numerals in the attached figures: 21. Acquisition module; 22. Processing module; 31. Processor; 32. Communication bus; 33. User interface; 34. Network interface; 35. Memory. Detailed Implementation

[0014] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0015] In the description of the embodiments of this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a specific manner.

[0016] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0017] To address the aforementioned technical problems, this application provides a laser ultrasonic testing method for metal grain size under high-temperature dynamic conditions, referring to... Figure 1 , Figure 1 This is a flowchart illustrating a high-temperature dynamic metal grain size laser ultrasonic testing method provided in an embodiment of this application. The method is applied to a server and includes steps S110 to S160, as follows:

[0018] S110. A programmable high-temperature environment is constructed using a heating furnace with a transparent quartz window, a temperature controller, and an infrared thermometer. The programmable high-temperature environment is used to control the heating and holding of at least one metal sample clamped in the heating furnace, and to obtain temperature condition information corresponding to a preset detection area on the surface of the metal sample.

[0019] Specifically, the high-temperature resistant sample holder can be connected to the server via wired or wireless means, meaning the server controls the automatic clamping of the high-temperature resistant sample holder. When clamping a metal sample into the furnace chamber using a high-temperature resistant sample holder, the positioning reference of the high-temperature resistant sample holder is first used as a unified clamping reference to limit and lock the metal sample, ensuring that the metal sample maintains a fixed spatial relationship with the positioning reference during heating, heat preservation, and subsequent dynamic conditions. After clamping, the orientation and height of the metal sample are finely adjusted with a preset detection area as the target, so that the preset detection area is within the effective transmission range of the transparent quartz window and consistent with the incident direction of the external light path, thereby forming an unobstructed channel between the transparent quartz window and the preset detection area in the laser propagation direction. In this process, the effective transmission range of the transparent quartz window is limited to the beam envelope area that the excitation spot and the probe spot may cover within the entire stroke of the two-dimensional moving platform. By checking the margin of the beam envelope area within the boundary of the transparent quartz window, subsequent thermally induced refractive index changes or minor mechanical offsets can prevent the beam from sweeping to the window boundary, thus introducing energy attenuation or scattering interference.

[0020] When temperature control parameters are pre-written into the temperature controller, the heating rate, segmented temperature points, temperature range, holding time, and allowable temperature fluctuation threshold are solidified into an executable temperature program. The temperature controller then uses this temperature program as the target to perform closed-loop adjustment of the heating power of the furnace. During the heating phase, the heating rate gradually approaches the segmented temperature points, and automatically switches to a slow-rise or fine-tuning mode when approaching the temperature range boundary to reduce the impact of temperature overshoot on the temperature stability of the preset detection area. During the holding phase, the heating power is finely modulated according to the allowable temperature fluctuation threshold to suppress temperature drift. At the same time, the stage identifiers of the temperature program are synchronously output to the industrial control computer, so that the subsequent temperature measurement sequence can correspond one-to-one with the temperature control state, thereby establishing a traceable link between the formation of temperature condition information and the execution process of temperature control parameters.

[0021] When setting the emissivity parameter for the infrared thermometer and the transmittance correction parameter for the transparent quartz window, the emissivity parameter is first set to a value consistent with the surface condition of the metal sample. This value is then kept consistent with the metal sample material grade, surface roughness, oxide film condition, and temperature measurement band settings, ensuring that the temperature measurement value output by the infrared thermometer reflects the true radiation characteristics of the metal sample surface. Next, a transmittance correction parameter is configured for the transparent quartz window, compensating for the transmission loss of the transparent quartz window within the temperature measurement band when calculating the temperature measurement value. The transmittance correction parameter is then correlated with the thickness of the transparent quartz window, material batch, and transmittance variation trend at high temperatures, thereby reducing the systematic temperature measurement deviation introduced by the transparent quartz window. To avoid contamination of the temperature measurement field of view by thermal radiation background, the temperature measurement field of view of the infrared thermometer is limited to covering a preset detection area or a temperature observation area with a fixed geometric offset from the preset detection area. This geometric offset is then fixed under the clamping reference to ensure a stable correspondence between the temperature measurement sequence and the subsequent detection point distribution.

[0022] During the temperature measurement sequence acquisition and time alignment in the heating and heat preservation control processes, the infrared thermometer outputs temperature measurement values ​​at a fixed sampling period, and the industrial control computer adds a unified timestamp to each temperature measurement value. Simultaneously, it reads the temperature control status corresponding to the temperature control parameters from the temperature controller and adds the same timestamp system to the temperature control status, thereby matching the temperature measurement sequence and the temperature control status under the same detection time axis. When there are multiple metal samples, the sampling position of the temperature measurement sequence is stably mapped to the preset detection area of ​​each metal sample, and the temperature measurement sequence is bound to the sample index to form temperature condition information that can characterize "sample-preset detection area-time axis-temperature control status". This allows each synchronous trigger of subsequent laser ultrasonic testing to query the corresponding temperature condition information under the same timestamp and complete the condition labeling.

[0023] When the temperature measurement sequence meets the preset temperature stability criterion and a stable temperature range is marked, the temperature stability criterion is defined as the fluctuation amplitude and trend of the temperature measurement sequence within the sliding time window simultaneously meeting a threshold constraint. The industrial control computer performs real-time determination on the temperature measurement sequence to output the stable temperature range, thereby limiting subsequent laser ultrasonic testing to only be performed within the stable temperature range, ensuring consistency between temperature operating conditions and high-temperature dynamic operating conditions. The temperature stability criterion can be expressed as:

[0024]

[0025] Wherein, T(t) represents the temperature measurement value output by the infrared thermometer at time t, the value of which is obtained by sampling by the infrared thermometer and has been corrected by emissivity and transmittance parameters; The starting time of the sliding time window is indicated by the industrial control computer, which advances it step by step according to the detection time axis; W represents the width of the sliding time window, and its value is set according to the thermal inertia of the heating furnace and the time resolution of the target detection. This indicates the allowable temperature fluctuation threshold, which can be the same as the allowable temperature fluctuation threshold in the temperature control parameters or set according to stricter access conditions. It represents the rate of change of the temperature measurement value over time, and its value is approximately obtained by the difference between adjacent sampling times in the temperature measurement sequence; The threshold value represents the rate of temperature change, used to constrain the temperature to be in a slow or near-steady state. The value is determined based on the sensitivity of the material grain evolution to temperature disturbances and the length of the subsequent laser-ultrasound acquisition window. When the above two constraints are satisfied simultaneously within the same sliding time window, the time interval corresponding to the sliding time window is merged into the temperature stability interval, and the temperature stability interval is bound to the subsequent synchronous triggering allowable triggering conditions, so that the laser-ultrasound detection data naturally carries a traceable temperature stability guarantee.

[0026] S120: Control the linear motor to drive the heating furnace and multiple metal samples inside to move in a uniform reciprocating motion along a preset motion direction to form a controllable and repeatable dynamic working condition, and simultaneously acquire the motion position and speed of the metal samples during the detection process as speed working condition information.

[0027] Specifically, the linear motor can be connected to the server via wired or wireless means, allowing the server to control the linear motor for automatic operation as a drive source. When the guide rail direction of the linear motor's moving mechanism is set to the preset motion direction, the geometric reference surface of the heating furnace shell and the mounting reference surface of the linear motor's moving mechanism are used as assembly references to complete the rigid connection between the heating furnace and the linear motor's moving mechanism. The guide rail direction is defined as the preset motion direction. Simultaneously, the distribution range of the preset detection area of ​​multiple metal samples in the preset motion direction is used as the basis for stroke design to verify whether the effective stroke of the linear motor in the preset motion direction covers the effective scanning length of all preset detection areas. During the verification process, the effective scanning length is limited to the projection range of the preset detection area that the excitation spot and the probe spot may cover during the scanning process of the two-dimensional moving platform. A safety margin is reserved for the reversing section and the endpoint buffer to ensure that the stable motion window of the uniform speed section can completely cover the effective scanning length and avoid speed fluctuations near the endpoints from affecting the consistency of detection point coverage.

[0028] When controlling the high-temperature resistant sample fixture to simultaneously fix multiple metal samples in the heating furnace, each metal sample is first positioned using the positioning reference of the high-temperature resistant sample fixture. The clamping distance of each metal sample is then consistently constrained along the preset movement direction, ensuring that the preset detection area of ​​each metal sample maintains a fixed spatial relationship with the positioning reference in the preset movement direction. Subsequently, a locking structure rigidly constrains the metal sample to the high-temperature resistant sample fixture, preventing relative displacement between the metal sample and the heating furnace as it moves with the linear motor. This allows the speed information to simultaneously represent the actual movement state of both the heating furnace and the metal sample. To ensure the traceability of subsequent position markings, the sample index of each metal sample and the start and end position intervals of its preset detection area in the preset movement direction are registered. This allows the subsequent real-time movement position to be mapped to the sample index and the spatial index of the corresponding preset detection area, thereby supporting sample differentiation and detection point positioning in batch testing.

[0029] When the industrial control computer writes motion control parameters for uniform reciprocating motion to the linear motor, the reciprocating motion trajectory is defined as a periodic motion including acceleration, uniform speed, deceleration, and reversing segments. The motion control parameters are then written into the trajectory planning interface of the linear motor controller, enabling the linear motor to output drive commands according to the trajectory planning in a closed-loop position control manner. Specifically, by setting the acceleration constraints of the acceleration and deceleration segments to values ​​that match the overall mass of the heating furnace, load inertia, and guide rail friction characteristics, the linear motor can enter the uniform speed segment within a controllable acceleration distance and exit it within a controllable deceleration distance. Simultaneously, by setting the endpoint position, endpoint dwell time, or endpoint transition speed constraints of the reversing segment, the dynamic impact of the reciprocating reversing process is limited to the endpoint buffer zone, thus forming a stable motion window with consistent time and position within each reciprocating cycle. To ensure the alignment consistency between the stable motion window and the subsequent synchronous trigger acquisition window, the start and end positions of the stable motion window are calibrated using a preset position threshold in the motion direction, and this position threshold is fixed in the industrial control computer as a spatial criterion for allowing triggering, ensuring that subsequent detection is only performed within the stable motion window.

[0030] When acquiring real-time motion position and generating motion speed feedback information within a stable motion window, the position sensor configured on the linear motor continuously outputs the real-time motion position of the heating furnace in the preset motion direction, and the industrial control computer adds a unified timestamp to each real-time motion position to form a motion position sequence; based on the real-time motion position at adjacent sampling times, motion speed feedback information is calculated, and the industrial control computer also adds a unified timestamp to the motion speed feedback information to form a motion speed sequence, so that the motion position sequence and the motion speed sequence maintain a one-to-one correspondence on the same time axis, thereby constructing speed condition information; the motion speed feedback information can be obtained by differential approximation and expressed as:

[0031] in, Indicates the timestamp The corresponding motion speed feedback information, whose positive or negative sign is used to represent the direction of reciprocating motion; Indicates the location sensor at the timestamp The output real-time motion position is obtained by sampling from the position sensor and has been calibrated in the linear motor coordinate system; Indicates the position sensor's last sampling timestamp Output the real-time motion position; The time interval between two adjacent samples is represented by the unified timestamp system of the industrial control computer. Based on the motion position sequence and motion speed sequence, the time interval within the stable motion window that meets the speed fluctuation threshold is further marked as the uniform speed stable interval. This makes the speed condition information not only include real-time motion position and motion speed feedback information, but also includes the uniform speed stable interval identifier that can be used for subsequent detection access control and sample labeling, thereby realizing the controllability, repeatability and traceability of dynamic conditions.

[0032] S130: The two-dimensional moving platform is controlled to coordinate the position of the pulsed laser excitation head and the laser interferometer receiving head, so that the excitation spot and the detection spot always cover the preset scanning area during the high-temperature dynamic movement of the metal sample, thereby forming multiple stable detection points in the same detection process and realizing in-situ scanning detection of multiple samples.

[0033] Specifically, the two-dimensional mobile platform can be connected to the server via wired or wireless means, allowing the server to control the platform for automatic position adjustment. When fixing the two-dimensional mobile platform to a rigid base outside the heating furnace, the mounting reference surface of the heating furnace shell is first precisely aligned with the mounting reference surface of the rigid base. A thermally insulated connection structure that resists thermal drift forms a stable mechanical closed loop between the two-dimensional mobile platform and the rigid base, ensuring the platform maintains stability even under vibrations introduced by the heating furnace's heating and the reciprocating motion of the linear motor. Subsequently, a first and second orthogonal movement axis are defined on the two-dimensional mobile platform, and the origin of the coordinates of the first and second movement axes is calibrated through stroke limiting and zero-point reset processes, ensuring the two-dimensional mobile platform remains stable in planar coordinates. The system has a traceable positioning reference. Based on this, the pulsed laser excitation head and the laser interferometer receiver head are mounted together on the same carrier frame of the two-dimensional moving platform. The positioning holes, positioning pins and fine-tuning mechanism of the carrier frame define the relative installation reference between the excitation spot and the detection spot, so that the excitation spot and the detection spot maintain a constant relative offset relationship throughout the entire stroke of the two-dimensional moving platform. Moreover, this relative offset relationship can be repeatedly reproduced at any coordinate position of the two-dimensional moving platform, thereby avoiding relative drift between the excitation spot and the detection spot during the scanning process and laying a consistent geometric constraint for subsequent detection of the same point or the same neighborhood.

[0034] After the metal sample is clamped, when establishing the mapping relationship between the preset scanning area and the plane coordinate system, the preset scanning area of ​​the metal sample is first defined as a two-dimensional region with boundary reference and grid reference in the metal sample coordinate system. This metal sample coordinate system is then bound to the positioning reference of the high-temperature resistant sample fixture, so that the preset scanning area has a fixed spatial position in the heating furnace coordinate system. Subsequently, the geometric reference of the metal sample is externally calibrated through a transparent quartz window. The geometric reference includes the edge of the metal sample, the boundary mark of the preset scanning area, or the preset positioning hole mark. The two-dimensional moving platform is controlled to move to several calibration points, so that the excitation spot falls sequentially on the position corresponding to the geometric reference to collect the coordinate value of the calibration point in the plane coordinate system of the two-dimensional moving platform. This establishes the coordinate representation of the preset scanning area in the plane coordinate system of the two-dimensional moving platform. To ensure the stability of the mapping relationship under high temperature conditions, the calibration process is performed within the temperature stability range, and the temperature condition information corresponding to the calibration time is recorded as the condition reference of the mapping relationship. This allows the effectiveness of the mapping relationship to be maintained by a small correction if a thermally induced change in refractive index causes an optical path deviation.

[0035] After mapping the preset scanning area to the two-dimensional moving platform's plane coordinate system, when using the mapping relationship as the control reference for the two-dimensional moving platform to execute scanning motion, the boundary constraints and scanning grid definitions of the preset scanning area are first solidified in the two-dimensional moving platform controller. The scanning grid definitions are then converted into target coordinate sequences of the first and second moving axes of the two-dimensional moving platform, ensuring that each target coordinate corresponds to a detection point and the spacing between detection points meets the preset spatial resolution requirements. When the preset scanning area adopts a row scanning mode, the first moving axis is defined as the row scanning direction, and the second moving axis is defined as the inter-row stepping direction. The movement range of the row scanning direction completely covers the boundary range of the preset scanning area, and the step distance of the inter-row stepping direction can cover all rows of the preset scanning area without missing the scanning grid. At the same time, the target coordinates of each detection point are bound to the detection point index, thereby ensuring that the subsequent laser ultrasonic detection results can be traced back to the two-dimensional moving platform coordinates according to the detection point index and further mapped to the spatial position within the preset scanning area of ​​the metal sample.

[0036] When controlling the two-dimensional moving platform to perform coordinated displacement according to the preset scanning area to form multiple stable detection points, the two-dimensional moving platform moves sequentially according to the target coordinate sequence in the directions of the first and second moving axes, and enters a position stabilization state after reaching each target coordinate. The position stabilization state is constrained by both the positioning error threshold and the dwell time threshold to ensure that the instantaneous micro-vibrations of the support frame, the pulsed laser excitation head, and the laser interferometer receiver head at the target coordinate are attenuated to an acceptable range. In the position stabilization state, the excitation spot and the detection spot remain within the preset scanning area during the high-temperature dynamic movement of the metal sample with the heating furnace. This is achieved by coordinating the target coordinate sequence of the two-dimensional moving platform with the stabilization of the linear motor. The motion window is time-coordinated so that the trigger time of the detection point falls within the overlapping acquisition window of the uniform velocity stable range and the temperature stable range. This ensures that each detection point is not only spatially stable and repeatable, but also has consistent operating condition labeling with temperature and velocity operating condition information. If online constraints are required on the landing point deviation between the excitation spot and the probe spot, a landing point deviation criterion is constructed based on the signal quality indication received by the laser interferometer. When the signal quality indication deviates from the preset threshold, the two-dimensional moving platform is controlled to perform a micro-compensation displacement to restore the landing point coupling quality. The micro-compensation displacement is included in the actual coordinate record of the detection point, thereby maintaining the repeatability and traceability of the detection point without changing the preset scanning area definition.

[0037] When switching coordinates during in-situ scanning inspection of multiple metal samples, the clamping distance of each metal sample along the preset movement direction and its preset scanning area boundary reference are first registered during the clamping stage. This ensures that each metal sample corresponds to a preset scanning area coordinate range in the two-dimensional moving platform plane coordinate system. A mapping table between the sample index and the preset scanning area coordinate range is established in the industrial control computer. After the two-dimensional moving platform completes the target coordinate sequence corresponding to the preset scanning area of ​​the first metal sample, the target coordinate starting point of the two-dimensional moving platform is switched to the preset scanning area starting coordinate of the next metal sample according to the mapping table. The scanning grid definition, row scanning direction, and inter-row stepping direction remain unchanged. Thus, the same scanning strategy can be reused to complete the in-situ scanning inspection of multiple metal samples without changing the scanning trajectory definition. To avoid conflicts with synchronous trigger acquisition during coordinate switching, coordinate switching is limited to execution within the non-acquisition window. The non-acquisition window is constrained by setting the allowed trigger state to the prohibited trigger state of the synchronous control card. This ensures that coordinate switching does not introduce invalid detection points or cause mismatch of detection point indexes, thereby ensuring the consistency of spatial coordinates, detection point indexes, and working condition annotations in the batch inspection of multiple samples.

[0038] S140. After the temperature and speed operating conditions information reach a stable state, the control synchronization card simultaneously triggers the nanosecond pulse laser to excite broadband ultrasonic waves on the surface of the metal sample, and triggers the laser interferometer to receive the surface micro-vibration echo signal caused by ultrasonic propagation in a non-contact manner. At the same time, the high-speed data acquisition card is controlled to synchronously acquire and transmit the ultrasonic time series, temperature operating conditions information, and speed operating conditions information to the industrial control computer for storage and processing, so as to obtain the laser ultrasonic detection signal containing multiple echoes and the corresponding operating conditions information.

[0039] Specifically, after the temperature and speed operating conditions reach a stable state, when a joint judgment is made and a trigger-allowing command is issued, the industrial control computer continuously reads the temperature stability interval marker in the temperature operating condition information and the uniform speed stability interval marker in the speed operating condition information. Under a unified timestamp system, it performs a parallel constraint judgment on whether the current detection time falls within both the temperature stability interval and the uniform speed stability interval. If the judgment is successful, the trigger-allowing command is written to the trigger enable register of the synchronization control card, and the time, detection point index, and operating condition index of this trigger-allowing command are recorded synchronously. This ensures that subsequent triggering behaviors are strictly bound to stable operating conditions at the data level. The joint judgment can be described in the form of an indicator function:

[0040] Where E(t) represents the triggering determination result at time t, with a value of 1 indicating that triggering is allowed and a value of 0 indicating that triggering is prohibited; This indicates an indicator function that takes the value 1 when the condition inside the parentheses is true, and 0 otherwise. This represents the set of temperature stability intervals determined by temperature operating condition information, which consists of the time periods in which the temperature measurement sequence satisfies the temperature stability criterion. This represents the set of uniformly stable intervals determined by speed condition information, consisting of time periods during which the motion speed feedback information satisfies the speed stability criterion; when At this time, the industrial control computer sends a trigger enable command to the synchronous control card and saves the trigger enable state along with the current detection point index to avoid false triggering caused by sudden changes in operating conditions.

[0041] When the synchronization control card serves as a unified timing reference source and outputs excitation, reception, and acquisition trigger signals, it enters a trigger-ready state upon receiving a trigger-allowed command. Upon the arrival of the trigger cycle, it simultaneously outputs three hardware trigger signals. The excitation trigger signal connects to the external trigger port of the nanosecond pulse laser to control the laser to release single-pulse energy at a preset detection point and excite broadband ultrasound. The reception trigger signal connects to the sampling gate port of the laser interferometer to ensure the interferometer enters a high-sensitivity reception state with a time reference aligned with the excitation trigger signal. The acquisition trigger signal connects to the trigger input port of the high-speed data acquisition card to define the start of the high-speed sampling window. To ensure that multiple echoes fall completely within the sampling window, a fixed delay relative to the excitation trigger signal is configured within the synchronization control card, and trigger jitter is constrained to ensure that excitation, reception, and acquisition are consistent and reproducible on the time axis. The trigger alignment relationship can be described by the following formula:

[0042] in, The output time of the trigger signal is defined by the reference clock of the synchronization control card; Indicates the output time when the trigger signal is received; Indicates the output time of the acquisition trigger signal; This represents the delay between the received trigger signal and the excitation trigger signal, and its value is obtained by calibrating the gating response time and optical path delay of the laser interferometer. This represents the delay between the acquisition trigger signal and the excitation trigger signal. Its value is set based on the expected arrival time of the first echo and the sampling window coverage strategy. By fixing this delay, the sampling window is opened at the same reference point after excitation each time, thereby ensuring the consistency between the echo threshold and the echo index in subsequent processing.

[0043] When the high-speed data acquisition card receives the acquisition trigger signal, forms an ultrasonic time series, and acquires operating condition information in parallel, the high-speed data acquisition card starts a high-speed sampling clock the instant the acquisition trigger signal arrives. Within a preset sampling window, it samples the ultrasonic echo signal output by the laser interferometer at equal intervals to form an ultrasonic time series. Simultaneously, it acquires temperature and velocity operating condition information in parallel through a multi-channel synchronous acquisition interface, ensuring that the ultrasonic time series, temperature and velocity operating condition information share the same trigger timestamp under the same trigger event. To ensure that the sampling window covers the first echo and subsequent echoes, the sampling window length and number of sampling points are configured according to the sampling rate and the upper limit of the echo propagation time, and the sampling time series is explicitly bound to the trigger timestamp. The sampling relationship can be described by the following formula:

[0044]

[0045] Wherein, s[n] represents the value of the nth sampling point of the ultrasonic time series, which is obtained by sampling the output electrical signal of the laser interferometer by the high-speed data acquisition card; u(t) represents the instantaneous value of the ultrasonic echo signal output by the laser interferometer at time t, which is obtained by photoelectric conversion of the surface micro-vibration echo signal; Indicates the start time of the sampling window, aligned with the acquisition trigger signal; The sampling period is defined as the reciprocal of the sampling rate and is determined by the internal sampling clock of the high-speed data acquisition card. N represents the number of sampling points, which is determined by the preset sampling window length and the sampling period, satisfying the condition that the sampling window length equals... t[n] represents the absolute sampling time corresponding to the nth sampling point, which is used to define the echo arrival time, echo threshold and multiple echo index in subsequent signal processing in a consistent manner; the temperature and speed operating conditions information acquired in parallel are written to the operating condition channel buffer with the same trigger timestamp under the same trigger event, so that the operating condition information can be one-to-one with the ultrasonic time series in the time dimension.

[0046] When transmitting a data frame containing ultrasonic time series, temperature condition information, speed condition information, and corresponding timestamps to the industrial control computer for storage and processing, the high-speed data acquisition card encapsulates the ultrasonic time series cache, condition information cache, and trigger timestamp into a single-frame data structure after completing one sampling window acquisition. This data structure also simultaneously writes metadata such as the detection point index, two-dimensional moving platform coordinates, real-time linear motor position, and allowed trigger determination results, ensuring that each data frame has traceable spatial positioning and condition labeling. Subsequently, the single-frame data structure is transmitted to the industrial control computer's receiving buffer via the high-speed bus. The industrial control computer performs integrity verification on the receiving buffer and then... The timestamps are sequentially written to persistent storage, and the single-frame data structure is registered to the subsequent processing queue for echo identification, time-domain attenuation feature extraction, and frequency-domain attenuation feature extraction. To ensure frame-level consistency during multiple trigger acquisitions under long-term high-temperature dynamic conditions, the unified timestamp of the data frame is generated using the reference clock of the synchronization control card or the clock of the industrial control computer phase-locked with the synchronization control card. This ensures that the trigger timestamp remains monotonically increasing throughout the entire process and can be used for cross-device alignment. Ultimately, this yields a laser ultrasonic detection signal containing multiple echoes and its corresponding temperature and speed condition information, providing a reproducible data foundation for subsequent grain size inversion model training and inference.

[0047] S150. Using a pre-prepared standard sample with the same grade as the metal material to be tested but different grain size and uniform grain distribution within the C-scan detection area, the laser ultrasonic in-situ detection process is reproduced under various motion speed conditions covering the range from room temperature to high temperature. This allows for the extraction of time-domain and frequency-domain attenuation features from the laser ultrasonic detection signal. Based on the time-domain and frequency-domain attenuation features and the corresponding working condition information, a high-dimensional feature vector is constructed. Combined with the grain size supervision label of the standard sample obtained in a preset manner, an artificial intelligence model of ultrasonic attenuation and grain size under high-temperature dynamic working conditions is trained.

[0048] Specifically, when preparing standard samples and uniformly defining the C-scan detection area, raw materials of the same grade as the metal material to be tested are first selected. Multiple sets of standard samples with different grain sizes are prepared through heat treatment or recrystallization control processes. A unique sample index is established for each standard sample to be used throughout the entire process of subsequent data, labeling, and model training. Then, the spatial range of the C-scan detection area is defined on the surface of each standard sample using the same boundary reference. A set of grid points is generated using a uniform scanning grid density to ensure that the number of grid points, grid point spacing, and boundary alignment rules of different standard samples in the C-scan detection area are consistent, thereby ensuring that the laser ultrasonic detection signals of different standard samples can be directly aligned in the spatial sampling structure.

[0049] When performing laser ultrasonic C-scan detection under static conditions at room temperature and screening for grain distribution uniformity based on spatial distribution consistency, each standard sample is kept at room temperature and in a static state, and is arranged according to a grid point set. The laser ultrasonic testing signals are sequentially triggered to obtain the laser ultrasonic testing signal and its time-domain or frequency-domain attenuation characterization quantity corresponding to each grid point. To transform "spatial distribution consistency" into a quantifiable index, a uniform attenuation scalar is first extracted for each grid point as a spatial consistency evaluation quantity. Then, the coefficient of variation of the discrete distribution of this attenuation scalar within the C-scan detection area is calculated and compared with a uniformity threshold to screen out target standard samples that meet the uniformity requirements. The coefficient of variation can be expressed as:

[0050] in, The coefficient of variation represents spatial uniformity; a smaller value indicates a more uniform spatial distribution. M represents the number of effective grid points within the C-scan detection area, and its value is equal to... The quantity after deducting invalid points; This represents the attenuation scalar of the m-th effective grid point, whose value is extracted from the laser ultrasonic detection signal of that grid point according to a uniform rule; This represents the mean of the attenuation scalars at all valid grid points, and its value is determined by... Get; when The corresponding standard sample is then identified as the target standard sample and proceeds to the subsequent modeling process. This represents the uniformity threshold, and its value is set based on the allowable degree of spatial non-uniformity and the representativeness requirements of subsequent monitoring labels.

[0051] When reproducing the in-situ laser ultrasonic testing process under various motion speeds and covering a range from room temperature to high temperature, and acquiring the target laser ultrasonic testing signal within a stable operating condition range, a combination of temperature and speed setpoints is sequentially set for each target standard sample. This allows the heating furnace to generate temperature operating condition information and mark the temperature stable range under the target temperature setpoint, and the linear motor to generate speed operating condition information and mark the uniform speed stable range under the target speed setpoint. The overlapping area between the temperature stable range and the uniform speed stable range is then taken as the allowable acquisition range. Within the allowable acquisition range, data is collected using a grid system. The system locates and triggers data acquisition, ensuring that each target laser ultrasonic detection signal is bound to the temperature and speed conditions under the same trigger timestamp, thereby writing high temperature and dynamic conditions into the sample as explicit variables.

[0052] When preprocessing the target laser ultrasonic detection signal, identifying multiple echoes, and extracting time-domain and frequency-domain attenuation features, the process first performs time alignment on each target laser ultrasonic detection signal to reduce trigger jitter. Then, bandpass filtering and wavelet denoising are performed to suppress ineffective frequency band noise and non-stationary interference. Subsequently, the first echo and subsequent echoes are detected within a unified propagation time window, and an echo index set is established, allowing each echo to be aligned across samples under the same echo index. Time-domain attenuation features can be characterized by the peak-to-envelope ratio or energy ratio of adjacent echoes, organized into a sequence using echo indices. Frequency-domain attenuation features can be characterized by the attenuation ratio of echo spectral energy across multiple frequency bands, also organized into a sequence using echo indices. For example, the peak-to-envelope ratio of the k-th echo... Constructing the attenuation of adjacent echoes:

[0053] in, This represents the time-domain attenuation of the k-th echo relative to the first echo, and its value increases as the attenuation increases. This represents the peak value of the envelope of the first echo, and its value is obtained from the maximum value of the envelope curve within the time window of the first echo. The peak value of the envelope of the k-th echo is obtained by taking the maximum value of the envelope curve within the time window of the k-th echo; k is the echo index, and its range is determined by the echo identification result; for the frequency domain attenuation characteristics, first analyze the k-th echo signal... The amplitude spectrum is obtained by performing a discrete Fourier transform within its time window. Then, calculate the band energy for the preset frequency band set and construct the band energy attenuation ratio:

[0054] in, This represents the frequency band energy of the k-th echo in the b-th preset frequency band; This represents the set of frequency indices corresponding to the b-th preset frequency band, whose values ​​are obtained by mapping the sampling rate to the upper and lower limits of the frequency band. q represents the complex spectral value of the k-th echo amplitude spectrum at frequency index r, and its value is obtained by discrete Fourier transform; k,b This represents the frequency domain attenuation of the k-th echo relative to the first echo in the b-th frequency band. This represents the frequency band energy of the first echo in the b-th frequency band. By extracting multiple k and multiple b and organizing them according to the echo index, we obtain the time-domain attenuation characteristics and frequency-domain attenuation characteristics that can simultaneously reflect multiple echo attenuations and spectral selective attenuation.

[0055] When binding operating condition information with time-domain and frequency-domain attenuation features and constructing a high-dimensional feature vector with a fixed field structure, the trigger timestamp is used as the primary key. Temperature and speed operating condition information corresponding one-to-one with the trigger event are appended to the same feature record. Simultaneously, the time-domain and frequency-domain attenuation feature sequences organized by echo index are expanded according to a fixed field order and written into the feature record. This ensures consistency in the number of fields, field order, and normalization strategy across all samples, thereby guaranteeing a fixed input dimension for the artificial intelligence model and comparability across operating conditions and samples. The high-dimensional feature vector can be formally represented as:

[0056] in, This represents the high-dimensional feature vector of a single detection point under a single triggering event. This represents the time-domain attenuation sequence from the 2nd echo to the Kth echo, where K is determined by the number of valid echoes obtained from echo identification. This represents the set of frequency domain attenuation from the 2nd echo to the Kth echo, and from the 1st frequency band to the Bth frequency band, where B is determined by the number of preset frequency bands; T represents the temperature measurement value in the temperature condition information bound to the trigger timestamp, whose value is output by the infrared thermometer and corrected for emissivity and transmittance; v represents the motion speed feedback information in the velocity condition information bound to the trigger timestamp, whose value is output by the position sensor and obtained through differential; x represents the motion position in the velocity condition information bound to the trigger timestamp, whose value is output by the position sensor and calibrated in the preset motion direction coordinate system; through this fixed field structure, each target laser ultrasonic detection signal carries a traceable condition label at the feature level and forms a consistent data structure with the echo attenuation characteristics.

[0057] When acquiring grain size supervision labels consistent with the C-scan detection area and training an artificial intelligence model, for each target standard sample, after completing multi-condition acquisition, sampling positions are selected according to the spatial range consistent with the C-scan detection area for metallographic or EBSD measurements. The measured grain size is aggregated into a grain size supervision label for the target standard sample in the C-scan detection area, establishing a one-to-one correspondence between the grain size supervision label and the high-dimensional feature vector set formed by the target standard sample in the C-scan detection area. During the training phase, the high-dimensional feature vector is used as input and the grain size supervision label is used as output. The training set and validation set are divided according to the sample index to avoid data leakage of the same target standard sample. The model parameters are solidified by minimizing the loss function between the predicted value and the supervision label, thereby obtaining an ultrasonic attenuation-grain size artificial intelligence model suitable for high-temperature dynamic conditions. In the embodiments of this application, the artificial intelligence model is preferably a Bayesian neural network model. Its core is that the weight parameters in the network are no longer regarded as deterministic fixed values, but are modeled as random variables with prior and posterior distributions, so that the model can explicitly quantify the prediction uncertainty while completing the grain size prediction. In this way, the model can not only output the predicted grain size, but also reflect the reliability of the predicted results under conditions such as high temperature dynamic conditions, signal noise fluctuations, and insufficient sample coverage.

[0058] For example, the loss function described above can be expressed as:

[0059] in, This represents the objective function used for training; The set of trainable parameters for the artificial intelligence model is represented, and its values ​​are updated through iterative optimization; N represents the number of samples in the training set, and its value is the number of high-dimensional feature vectors that enter the training set. This represents the grain size prediction of the artificial intelligence model for the nth high-dimensional feature vector, and its value is output by the model's forward inference. The grain size supervision label corresponding to the nth high-dimensional feature vector is taken by metallographic or EBSD measurement and is consistent with the spatial distribution of the C-scan detection area. represents the regularization coefficient, used to constrain model complexity; its value is determined by a combination of validation set performance and overfitting risk. The squared L2 norm of the parameter set is used to suppress overfitting caused by excessively large parameters. Through this training process, the model parameters converge under the constraints of samples covering the range from room temperature to high temperature and various motion speed conditions, thereby obtaining an artificial intelligence model that can realize grain size inversion using time-domain decay characteristics, frequency-domain decay characteristics and working condition information in high-temperature dynamic conditions.

[0060] S160. The laser ultrasonic acquisition and feature extraction process is carried out on the test sample under the target high temperature and target motion speed conditions. The obtained target high-dimensional feature vector is input into the artificial intelligence model, and the grain size prediction results and confidence intervals corresponding to each detection point are output. The grain size level is then calculated to realize the batch in-situ non-destructive quantitative detection of grain size of metal materials under high temperature dynamic conditions.

[0061] Specifically, when inputting the target high-dimensional feature vector into the artificial intelligence model according to the feature field order, normalization method, and data interface format determined during the training phase of the artificial intelligence model, and outputting the grain size prediction result, the industrial control computer first performs field consistency verification and field rearrangement on the target high-dimensional feature vector. This ensures that the feature fields corresponding to each detection point are strictly written into the model input tensor according to the field order fixed during the training phase. Simultaneously, during the normalization phase, three types of constraints are introduced: global statistical normalization, operating condition normalization, and domain alignment normalization, to suppress distribution drift caused by high-temperature dynamic operating conditions. Global statistical normalization is used to eliminate the magnitude differences between different feature fields; operating condition normalization is used to eliminate the systematic shift of temperature and speed operating conditions on the mean and variance of the feature fields; and domain alignment normalization is used to align the feature covariance of the test sample under the target operating condition with the feature covariance of the training set within the same metric space, thereby ensuring that the forward inference operation remains consistent with the training phase at the input distribution level. Normalization and domain alignment can be expressed as:

[0062]

[0063]

[0064] in, This represents the original value of the target high-dimensional feature vector of a single detection point. The value is obtained by concatenating time-domain attenuation features, frequency-domain attenuation features, temperature condition information, and speed condition information according to a fixed field structure. This represents the normalized and domain-aligned model input vector; T represents the temperature measurement value bound to the trigger timestamp of this detection point, which is output by the infrared thermometer and corrected for emissivity and transmittance parameters; v represents the motion velocity feedback information bound to the trigger timestamp of this detection point, which is output by the position sensor and obtained through differentiation. This represents the mean vector of operating conditions, used to characterize the systematic shift of features under different temperature and speed conditions; This represents the baseline mean vector of the training set, whose values ​​are statistically determined and fixed during the training phase. This represents the temperature basis function vector, whose values ​​are obtained by polynomial, piecewise, or spline expansion of the temperature measurements. The order of the expansion is constrained by both the fitting error and the risk of overfitting during the training phase. This represents the velocity basis function vector, whose values ​​are obtained by polynomial, piecewise, or spline expansion of the velocity feedback information. This represents the tensor product, used to explicitly represent the coupling terms of temperature and velocity; , , The coefficient matrix represents the mean model of the working conditions. Its values ​​are obtained and fixed during the training phase by fitting the working condition residuals of the training set to minimize them. This represents the operating condition scale matrix, used to perform operating condition-related scale normalization on each feature field. The baseline scale vector represents the training set, and its value is the field scale obtained from statistics during the training phase and then fixed. This indicates squaring by elements; , , The coefficient matrix represents the scaling model, and its values ​​are obtained and fixed during the training phase. This represents a numerically stable term, used to avoid numerical instability caused by the scale approaching zero. Its value is a preset positive decimal.

[0065]

[0066] in, This represents the linear transformation matrix used for feature domain alignment, which achieves covariance structure consistency by whitening and recoloring the second-order statistics of features in the training and test domains. It represents the covariance matrix of the training set in the normalized feature space, which is obtained and fixed by statistical analysis of the high-dimensional feature vectors during the training phase; represents the covariance matrix of the test sample in the same feature space, which is obtained by batch statistical analysis of the target high-dimensional eigenvectors under target operating conditions; η represents the covariance regularization coefficient, used to enhance numerical stability when the eigenvalues ​​of the covariance matrix are small or ill-conditioned, and its value is a preset positive decimal; I represents the identity matrix with the same dimension as the covariance matrix; exponent The exponent represents the inverse square root operation of a matrix, used for whitening features in the training domain. The square root operation of a matrix is ​​used to map whitened features to the covariance structure of the test domain, thereby achieving alignment of feature distributions in a second-order statistical sense.

[0067] During model inference, when generating confidence intervals and binding and storing them with the detection point index, 2D moving platform coordinates, and operating condition identifiers, the industrial control computer decomposes uncertainty into model uncertainty and observation noise uncertainty, and uses multiple stochastic inferences to form a prediction distribution. Then, it constructs confidence intervals using the quantiles of the prediction distribution. Simultaneously, it introduces the propagation uncertainties of temperature and speed operating condition information to reflect the impact of operating condition measurement fluctuations on the prediction. Multiple stochastic inferences can be achieved by enabling random deactivation, random sub-networks, or ensemble sub-models during the inference phase, ensuring that the same detection point operates under the same conditions. Given multiple sets of predicted samples, we then weight and aggregate these samples to obtain the predicted mean, predicted variance, and confidence interval. The predicted distribution and confidence interval can be expressed as follows:

[0068]

[0069]

[0070]

[0071] in, This represents the grain size prediction sample from the r-th random inference, with values ​​determined by the artificial intelligence model using a random mechanism. Below The output is obtained by performing forward inference; This represents the forward inference function of an artificial intelligence model; This represents the fixed set of model parameters, whose values ​​are fixed after convergence during the training phase. The value of represents the random state of the r-th inference, and is generated by random inactivation mask, random subnetwork selection, or ensemble model index; R represents the number of random inferences, and is set by the target confidence interval stability requirement and the computing power constraint of the industrial control computer. This represents the weight of the r-th inference, and its value can be set as follows: To form equal-weighted aggregation, or to adaptively set and satisfy inference quality indicators. ; This represents the center value of the predicted grain size. It represents the model uncertainty variance, reflecting the prediction dispersion caused by model parameters and structure; This represents the gradient vector of the predicted center value with respect to temperature and speed conditions. Its value is obtained by numerical sensitivity estimation of the model output in the dimension of condition variables or by the built-in differentiable structure of the model. The joint covariance matrix representing temperature and speed information is calculated from the statistical fluctuations of the temperature measurement sequence and motion speed feedback information within the stable range. This represents the variance of the uncertainty propagation under operating conditions, used to characterize the impact of fluctuations in operating condition measurements on grain size prediction. This represents the variance of the observation noise, and its value can be obtained by fixing the residual variance during the training phase or by adaptive estimation from the signal quality index. This represents the variance of the total forecast uncertainty; This represents the confidence interval for the predicted grain size. This represents the p-quantile operator for the predicted sample set, whose value is obtained by... The result is obtained by sorting and then taking the corresponding percentile position. The significance level is indicated by a pre-set confidence level. After the grain size prediction results and confidence intervals are completed, the industrial control computer binds them with the detection point index, the coordinates of the two-dimensional moving platform, and the condition identifier composed of temperature and speed condition information. The results are written into the structured storage with the trigger timestamp as the key, so that the output of each detection point has spatial traceability, condition traceability, and uncertainty interpretability.

[0072] After obtaining the grain size prediction result, when converting the grain size level according to the pre-built standard mapping rules and outputting the uncertainty range of the grain size level, the industrial control computer calls the standard mapping rule library to map the grain size prediction result to the center value of the grain size level. The upper and lower bounds of the grain size prediction confidence interval are also mapped to the lower and upper bounds of the grain size level under the same mapping rule library. Simultaneously, to avoid grade jumps caused by the grain size prediction result being near the grade boundary, a probability mapping with boundary softening is used to generate a grade distribution, and the uncertainty range is labeled by the grade distribution. This allows the grain size level to simultaneously reflect the center level and the reliable fluctuation range in engineering expression. The standard mapping and uncertainty range labeling can be represented as:

[0073]

[0074]

[0075]

[0076] Where G represents the random variable of grain size level; This represents the set of grain size levels allowed by the standard mapping rule base, and its values ​​are fixed by the range of standard levels adopted. This indicates the predicted center value of the grain size. With respect to total forecast uncertainty variance The probability that the grain size class is k under constraints; This represents the grain size representative value or interval center value corresponding to level k in the standard mapping rule base, and its value is pre-built into the standard mapping rule base. This represents the boundary softening coefficient, used to suppress numerical instability and excessive jumps at the level boundary. It takes a preset positive decimal value and can be adaptively adjusted according to the signal quality indication. This represents the center estimate of the grain size class, and its value is the expectation of the class probability distribution; This represents the range of uncertainty in the grain size classification, and its value is obtained by accumulating the probabilities of the classification probability distribution. and The corresponding level and position are determined; and These represent the lower and upper bounds of the grain size prediction confidence interval, respectively, and their values ​​are obtained from the aforementioned quantile confidence intervals. This represents a standard mapping function that takes a grain size value as input and outputs the corresponding discrete grain size level, the value of which is defined by the segmented interval rules of the standard mapping rule base; k, , These represent the discrete grade results obtained by predicting the center value and upper and lower bounds of the confidence interval for grain size according to the standard mapping function, respectively, and are used to verify consistency with the probability mapping output. The industrial control computer finally binds the grain size prediction results, confidence intervals, grain size grade center estimates and grain size grade uncertainty ranges with the detection point index, two-dimensional moving platform coordinates and working condition identifiers and writes them into the final output record. In the batch in-situ detection, the same inference, statistics and mapping logic is executed cyclically on all detection points to ensure that the quantitative detection results of grain size under high temperature dynamic working conditions are consistent in spatial dimension, working condition dimension and standard expression dimension.

[0077] This application also provides a high-temperature dynamic metal grain size laser ultrasonic testing system, referring to... Figure 2 , Figure 2This is a schematic diagram of a high-temperature dynamic metal grain size laser ultrasonic testing system provided in an embodiment of this application. The system is a server, which includes an acquisition module 21 and a processing module 22. The acquisition module 21 is used to construct a programmable high-temperature environment using a heating furnace with a transparent quartz window, a temperature controller, and an infrared thermometer. Based on this programmable high-temperature environment, it performs heating and holding control on at least one metal sample clamped in the heating furnace and acquires temperature condition information corresponding to a preset detection area on the surface of the metal sample. The acquisition module 21 is also used to control a linear motor to drive the heating furnace and multiple metal samples inside to reciprocate at a uniform speed along a preset motion direction, thereby forming a process... The system provides controllable and repeatable dynamic operating conditions, and simultaneously acquires the movement position and speed of the metal sample during the detection process as speed operating condition information; the processing module 22 is used to control the two-dimensional moving platform to coordinately adjust the positions of the pulsed laser excitation head and the laser interferometer receiving head, so that the excitation spot and the detection spot always cover the preset scanning area during the high-temperature dynamic movement of the metal sample, thereby forming multiple stable detection points in the same detection process and realizing in-situ scanning detection of multiple samples; the processing module 22 is also used to control the synchronous control card to simultaneously trigger the nanosecond pulsed laser to excite broadband ultrasonic waves on the surface of the metal sample after the temperature operating condition information and speed operating condition information enter a stable state, and trigger the synchronous control card to simultaneously trigger the nanosecond pulsed laser to excite broadband ultrasonic waves on the surface of the metal sample, and trigger the synchronous control card to detect the ultrasonic waves on the surface of the metal sample. The laser interferometer receives surface micro-vibration echo signals caused by ultrasonic propagation in a non-contact manner. Simultaneously, it controls a high-speed data acquisition card to synchronously acquire and transmit ultrasonic time-series, temperature, and speed information to an industrial control computer for storage and processing. This yields a laser ultrasonic detection signal containing multiple echoes and corresponding operating information. The processing module 22 is also used to reproduce the in-situ laser ultrasonic detection process using a pre-prepared standard sample of the same grade as the metal material to be tested but with different grain sizes and uniform grain distribution within the C-scan detection area. This process covers a range from room temperature to high temperature and various motion speeds, allowing for the extraction of time-domain and frequency-domain attenuation characteristics from the laser ultrasonic detection signal. Based on the time-domain attenuation characteristics, frequency-domain attenuation characteristics, and corresponding working condition information, a high-dimensional feature vector is constructed. Combined with the standard sample grain size supervision label obtained in a preset manner, an artificial intelligence model is trained to form an ultrasonic attenuation and grain size relationship under high-temperature dynamic working conditions. The processing module 22 is also used to perform laser ultrasonic acquisition and feature extraction on the test sample under target high temperature and target motion speed conditions. The obtained target high-dimensional feature vector is input into the artificial intelligence model, and the grain size prediction results and confidence intervals corresponding to each detection point are output. The grain size level is then calculated to realize batch in-situ non-destructive quantitative detection of grain size of metallic materials under high-temperature dynamic working conditions.

[0078] It should be noted that the above embodiments of the apparatus are only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0079] This application also provides an electronic device, with reference to... Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: at least one processor 31, at least one network interface 34, a user interface 33, a memory 35, and at least one communication bus 32.

[0080] The communication bus 32 is used to enable communication between these components.

[0081] The user interface 33 may include a display screen and a camera. Optionally, the user interface 33 may also include a standard wired interface and a wireless interface.

[0082] The network interface 34 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0083] The processor 31 may include one or more processing cores. The processor 31 connects to various parts of the server via various interfaces and lines, executing instructions, programs, code sets, or instruction sets stored in the memory 35, and calling data stored in the memory 35 to perform various server functions and process data. Optionally, the processor 31 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 31 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content to be displayed on the screen; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 31 and may be implemented as a separate chip.

[0084] The memory 35 may include random access memory (RAM) or read-only memory. Optionally, the memory 35 may include a non-transitory computer-readable storage medium. The memory 35 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 35 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 35 may also be at least one storage device located remotely from the aforementioned processor 31. Figure 3 As shown, the memory 35, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program for a high-temperature dynamic working condition metal grain size laser ultrasonic testing method.

[0085] exist Figure 3 In the electronic device shown, the user interface 33 is mainly used to provide an input interface for the user and to obtain the user input data; while the processor 31 can be used to call the application program stored in the memory 35 for a high-temperature dynamic working condition metal grain size laser ultrasonic detection method. When executed by one or more processors, the electronic device performs one or more methods as described in the above embodiments.

[0086] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0087] This application also provides a non-transitory computer-readable storage medium storing instructions. When executed by one or more processors, these instructions cause an electronic device to perform one or more of the methods described in the above embodiments.

[0088] The foregoing description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Those skilled in the art will readily conceive of other embodiments of this disclosure upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The specification and embodiments are considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

Claims

1. A laser ultrasonic testing method for metal grain size under high-temperature dynamic conditions, characterized in that, The method includes: A programmable high-temperature environment is constructed using a heating furnace with a transparent quartz window, a temperature controller, and an infrared thermometer. The programmable high-temperature environment is used to control the heating and holding of at least one metal sample clamped in the heating furnace, and to obtain temperature condition information corresponding to a preset detection area on the surface of the metal sample. The linear motor is controlled to drive the heating furnace and multiple metal samples inside to reciprocate at a constant speed along a preset motion direction, so as to form a controllable and repeatable dynamic working condition, and the motion position and speed of the metal samples during the detection process are acquired simultaneously as speed working condition information. The two-dimensional moving platform is controlled to coordinate the position of the pulsed laser excitation head and the laser interferometer receiving head, so that the excitation spot and the detection spot always cover the preset scanning area during the high-temperature dynamic movement of the metal sample, thereby forming multiple stable detection points in the same detection process and realizing in-situ scanning detection of multiple samples. After the temperature and speed operating conditions information reach a stable state, the control synchronization card simultaneously triggers the nanosecond pulse laser to excite broadband ultrasonic waves on the surface of the metal sample, and triggers the laser interferometer to receive the surface micro-vibration echo signal caused by ultrasonic propagation in a non-contact manner. At the same time, the high-speed data acquisition card is controlled to synchronously acquire and transmit the ultrasonic time series, the temperature and speed operating conditions information to the industrial control computer for storage and processing, so as to obtain the laser ultrasonic detection signal containing multiple echoes and the corresponding operating condition information. By using a pre-prepared standard sample with the same grade as the metal material to be tested but different grain size and uniform grain distribution within the C-scan detection area, the laser ultrasonic in-situ detection process is reproduced under various motion speed conditions covering the range from room temperature to high temperature. The time-domain attenuation features and frequency-domain attenuation features are extracted from the laser ultrasonic detection signal. Based on the time-domain attenuation features, the frequency-domain attenuation features and the corresponding working condition information, a high-dimensional feature vector is constructed. Combined with the standard sample grain size supervision label obtained in a preset manner, an artificial intelligence model of ultrasonic attenuation and grain size under high-temperature dynamic working conditions is trained. The laser ultrasonic acquisition and feature extraction process is carried out on the test sample under the target high temperature and target motion speed conditions. The obtained target high-dimensional feature vector is input into the artificial intelligence model, and the grain size prediction results and confidence intervals corresponding to each detection point are output. The grain size level is then calculated to realize the batch in-situ non-destructive quantitative detection of grain size of metallic materials under high temperature dynamic conditions.

2. The laser ultrasonic testing method for metal grain size under high-temperature dynamic conditions according to claim 1, characterized in that, The process involves constructing a programmable high-temperature environment using a heating furnace with a transparent quartz window, a temperature controller, and an infrared thermometer. This programmable high-temperature environment is used to control the heating and holding of at least one metal sample clamped within the heating furnace, and to acquire temperature information corresponding to a preset detection area on the surface of the metal sample. Specifically, this includes: The high-temperature resistant sample holder is controlled to clamp the metal sample in the furnace chamber of the heating furnace, and the preset detection area of ​​the metal sample and the positioning reference of the high-temperature resistant sample holder are kept in a fixed spatial relationship. At the same time, the transparent quartz window of the heating furnace and the preset detection area form an unobstructed channel in the laser propagation direction. The temperature controller is pre-programmed with temperature control parameters, including heating rate, segmented temperature points, temperature range, holding time, and allowable temperature fluctuation threshold, to control the heating furnace to perform closed-loop adjustment of heating power according to the temperature control parameters in order to form a high-temperature thermal field. The infrared thermometer is set with emissivity parameters consistent with those of the metal sample, and the transparent quartz window is set with transmittance correction parameters. During the heating and heat preservation control process, the temperature measurement sequence output by the infrared thermometer is collected, and the temperature measurement sequence is time-aligned with the temperature control state corresponding to the temperature control parameters to form temperature condition information consistent with the detection time axis. When the temperature measurement sequence meets the preset temperature stability criterion, the corresponding time interval is marked as the temperature stability interval, so as to limit the subsequent laser ultrasonic detection to be performed within the temperature stability interval, thereby ensuring the consistency between the temperature condition information and the high temperature dynamic condition.

3. The laser ultrasonic testing method for metal grain size under high-temperature dynamic conditions according to claim 1, characterized in that, The linear motor drives the heating furnace and multiple metal samples inside to reciprocate at a constant speed along a preset direction, forming a controllable and repeatable dynamic working condition. Simultaneously, the movement position and speed of the metal samples during the testing process are acquired as speed working condition information, specifically including: The direction of the guide rail of the linear motor's moving mechanism is set to the preset motion direction so that the movement stroke of the heating furnace in the preset motion direction covers the effective scanning length corresponding to the preset detection area of ​​multiple metal samples. The high-temperature resistant sample holder is controlled to simultaneously fix multiple metal samples in the heating furnace, and the preset detection area of ​​each metal sample is kept in a fixed spatial relationship with the positioning reference of the high-temperature resistant sample holder in the preset movement direction, so that when the linear motor drives the heating furnace to move, there is no relative displacement between each metal sample and the heating furnace. The industrial control computer writes motion control parameters for uniform reciprocating motion to the linear motor, so as to control the linear motor to perform closed-loop position control according to the reciprocating motion trajectory including acceleration segment, uniform speed segment, deceleration segment and reversal segment, and make the uniform speed segment form a stable motion window with consistent time and position in each reciprocating cycle. Within the stable motion window, the position sensor configured on the linear motor acquires the real-time motion position of the heating furnace in the preset motion direction, and generates corresponding motion speed feedback information based on the real-time motion position. By binding the real-time motion position and the motion speed feedback information with a unified timestamp, the speed condition information is constructed.

4. The laser ultrasonic testing method for metal grain size under high-temperature dynamic conditions according to claim 1, characterized in that, The control two-dimensional moving platform coordinates the positions of the pulsed laser excitation head and the laser interferometer receiving head, ensuring that the excitation spot and the detection spot always cover the preset scanning area during the high-temperature dynamic movement of the metal sample. This allows for the formation of multiple stable detection points and the realization of in-situ scanning detection of multiple samples during the same detection process. Specifically, this includes: The two-dimensional moving platform is fixed on a rigid base outside the heating furnace. The two-dimensional moving platform has a first moving axis and a second moving axis that are orthogonal to each other. The pulsed laser excitation head and the laser interferometer receiving head are mounted on the same support frame of the two-dimensional moving platform. The support frame defines the relative mounting reference between the excitation spot and the detection spot, so that the excitation spot and the detection spot maintain a constant relative offset relationship throughout the entire stroke of the two-dimensional moving platform. After the metal sample is clamped, the preset scanning area of ​​the metal sample is mapped to the plane coordinate system of the two-dimensional moving platform, and the mapping relationship is used as the control reference for the two-dimensional moving platform to perform scanning motion. The two-dimensional moving platform is controlled to perform coordinated displacement in the directions corresponding to the first moving axis and the second moving axis according to the preset scanning area, so that the excitation spot and the detection spot remain within the preset scanning area during the high-temperature dynamic movement of the metal sample with the heating furnace, so as to form multiple spatially stable and repeatable detection points in the same detection process. Based on the correspondence between the preset scanning areas of each metal sample in the plane coordinate system, the two-dimensional moving platform is controlled to switch coordinates between different preset scanning areas, so as to complete the in-situ scanning detection of multiple metal samples without changing the definition of the scanning trajectory.

5. The laser ultrasonic testing method for metal grain size under high-temperature dynamic conditions according to claim 1, characterized in that, After the temperature and velocity operating conditions information reach a stable state, the synchronous control card simultaneously triggers a nanosecond pulse laser to excite broadband ultrasonic waves on the surface of the metal sample, and triggers a laser interferometer to receive the surface micro-vibration echo signal caused by ultrasonic propagation in a non-contact manner. Simultaneously, the high-speed data acquisition card synchronously acquires and transmits the ultrasonic time series, the temperature and velocity operating conditions information to the industrial control computer for storage and processing, to obtain a laser ultrasonic detection signal containing multiple echoes and corresponding operating condition information, specifically including: After the temperature and speed operating conditions information reach a stable state, the industrial control computer is controlled to jointly determine the current detection time based on the temperature stable interval marker corresponding to the temperature operating conditions information and the uniform speed stable interval marker corresponding to the speed operating conditions information. When the temperature and speed operating conditions information simultaneously meet the stable interval conditions, an allow trigger command is sent to the synchronization control card. Using the synchronization control card as a unified timing reference source, after confirming that the synchronization control card has received the trigger permission command, the synchronization control card is controlled to output an excitation trigger signal to the nanosecond pulse laser, a reception trigger signal to the laser interferometer, and an acquisition trigger signal to the high-speed data acquisition card, so that the nanosecond pulse laser excites broadband ultrasonic waves at a preset detection point on the surface of the metal sample, and the laser interferometer receives the surface micro-vibration echo signal caused by ultrasonic propagation in a non-contact manner under the condition of being time-aligned with the excitation trigger signal; After confirming that the high-speed data acquisition card receives the acquisition trigger signal, it controls the high-speed data acquisition card to perform high-speed sampling of the ultrasonic echo signal output by the laser interferometer within a preset sampling window to form an ultrasonic time series covering the first echo and subsequent multiple echoes. At the same time, the temperature condition information and the speed condition information are acquired in parallel, and the same trigger timestamp is uniformly added to the ultrasonic time series, the temperature condition information and the speed condition information. The data frame containing the ultrasonic time sequence, temperature condition information, speed condition information and corresponding timestamps is transmitted as a whole to the industrial control computer for storage and processing to obtain a laser ultrasonic detection signal containing multiple echoes and corresponding condition information.

6. The laser ultrasonic testing method for metal grain size under high-temperature dynamic conditions according to claim 1, characterized in that, The process involves reproducing the in-situ laser ultrasonic detection process using a pre-prepared standard sample of the same grade as the metal material to be tested but with different grain sizes and uniform grain distribution within the C-scan detection area, under conditions covering room temperature to high temperature and various movement speeds. This process extracts time-domain and frequency-domain attenuation features from the laser ultrasonic detection signal. Based on the time-domain and frequency-domain attenuation features and the corresponding operating conditions, a high-dimensional feature vector is constructed. This vector is then combined with grain size monitoring labels of the standard sample obtained through a pre-defined method to train an artificial intelligence model relating ultrasonic attenuation to grain size under high-temperature dynamic conditions. Specifically, this includes: The standard sample is a set of multiple samples with the same grade as the metal material to be tested but different grain sizes. Each standard sample has a predefined consistent C-scan detection area so that the C-scan detection area is uniform in terms of spatial range, scanning grid density and boundary reference. Laser ultrasonic C-scan detection was performed on the C-scan detection area of ​​the standard sample under room temperature static conditions, and the uniformity of grain distribution was screened based on the consistency of the laser ultrasonic detection signal in spatial distribution. Target standard samples that meet the uniformity requirements were used for subsequent modeling. Under conditions covering room temperature to high temperature range and various motion speeds, the target standard sample is subjected to in-situ laser ultrasonic testing, and the target laser ultrasonic testing signal for modeling is acquired within the time interval when the working condition information is in a stable state. The target laser ultrasonic detection signal is preprocessed and multiple echoes are identified. Time-domain attenuation features characterizing ultrasonic energy attenuation behavior and frequency-domain attenuation features characterizing spectral change characteristics are extracted from the multiple echoes. The time-domain attenuation features and the frequency-domain attenuation features are then organized according to the echo index. The temperature and velocity information corresponding to the laser ultrasonic detection signal at the given time are bound to the time-domain attenuation feature and the frequency-domain attenuation feature to construct a high-dimensional feature vector with a fixed field structure. The grain size of the target standard sample, obtained by metallography or EBSD and consistent with the spatial characteristics of the C-scan detection area, is used as a supervision label and corresponds one-to-one with the high-dimensional feature vector to train an artificial intelligence model applicable to the relationship between ultrasonic attenuation and grain size under high-temperature dynamic conditions.

7. The laser ultrasonic testing method for metal grain size under high-temperature dynamic conditions according to claim 1, characterized in that, The process involves laser ultrasonic acquisition and feature extraction of the test sample under target high temperature and target motion speed conditions. The obtained high-dimensional feature vector is input into the artificial intelligence model, which outputs the grain size prediction results and confidence intervals corresponding to each detection point, and converts them into grain size grades. This enables batch in-situ non-destructive quantitative detection of grain size of metallic materials under high temperature dynamic conditions, specifically including: The target high-dimensional feature vector is input into the artificial intelligence model according to the feature field order, normalization method and data interface format determined during the training stage of the artificial intelligence model. The industrial control computer calls the solidified model parameters to perform forward inference operation, so as to output the grain size prediction result corresponding to each detection point. During the model inference process, statistical output information representing the uncertainty of prediction is acquired synchronously or statistical convergence processing is performed based on multiple inference results to generate a confidence interval corresponding to the grain size prediction result. The grain size prediction result and the confidence interval are then bound and stored with the detection point index, the two-dimensional moving platform coordinates, and the working condition identifier. After obtaining the grain size prediction result, the grain size prediction result is converted into the corresponding grain size level according to the pre-built standard mapping rule, and the confidence interval is converted into the uncertainty range label of the grain size level under the same mapping rule, so that the grain size level and uncertainty range can reflect the reliability of the prediction result and serve as the final output result of in-situ non-destructive quantitative detection of grain size under high temperature dynamic conditions.

8. A laser ultrasonic testing system for metal grain size under high-temperature dynamic operating conditions, characterized in that, The system is used to perform the high-temperature dynamic working condition metal grain size laser ultrasonic detection method as described in any one of claims 1 to 7, the system comprising an acquisition module and a processing module, wherein... The acquisition module is used to construct a programmable high-temperature environment through a heating furnace with a transparent quartz window, a temperature controller, and an infrared thermometer, so as to perform heating and heat preservation control on at least one metal sample clamped in the heating furnace according to the programmable high-temperature environment, and acquire temperature condition information corresponding to a preset detection area on the surface of the metal sample. The acquisition module is also used to control the linear motor to drive the heating furnace and the multiple metal samples inside to move in a uniform reciprocating motion along a preset motion direction, so as to form a controllable and repeatable dynamic working condition, and simultaneously acquire the motion position and motion speed of the metal sample during the detection process as speed working condition information. The processing module is used to control the two-dimensional moving platform to coordinate the position of the pulsed laser excitation head and the laser interferometer receiving head, so that the excitation spot and the detection spot always cover the preset scanning area during the high-temperature dynamic movement of the metal sample, thereby forming multiple stable detection points in the same detection process and realizing in-situ scanning detection of multiple samples. The processing module is also used to control the synchronous control card to simultaneously trigger the nanosecond pulse laser to excite broadband ultrasonic waves on the surface of the metal sample after the temperature and speed operating conditions information have entered a stable state, and to trigger the laser interferometer to receive the surface micro-vibration echo signal caused by ultrasonic propagation in a non-contact manner. At the same time, it controls the high-speed data acquisition card to synchronously acquire and transmit the ultrasonic time series, the temperature and speed operating conditions information to the industrial control computer for storage and processing, so as to obtain a laser ultrasonic detection signal containing multiple echoes and the corresponding operating condition information. The processing module is also used to reproduce the laser ultrasonic in-situ detection process under various motion speed conditions, using a pre-prepared standard sample with the same grade as the metal material to be detected but different grain size and uniform grain distribution within the C-scan detection area. This process extracts time-domain attenuation features and frequency-domain attenuation features from the laser ultrasonic detection signal. Based on the time-domain attenuation features, the frequency-domain attenuation features, and the corresponding working condition information, a high-dimensional feature vector is constructed. Combined with the standard sample grain size supervision label obtained in a preset manner, an artificial intelligence model is trained to form the relationship between ultrasonic attenuation and grain size under high-temperature dynamic working conditions. The processing module is also used to perform laser ultrasonic acquisition and feature extraction on the test sample under target high temperature and target motion speed conditions, input the obtained target high-dimensional feature vector into the artificial intelligence model, output the grain size prediction results and confidence intervals corresponding to each detection point, and convert them into grain size grades, so as to realize batch in-situ non-destructive quantitative detection of grain size of metal materials under high temperature dynamic conditions.

9. An electronic device, characterized in that, The electronic device includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions. The user interface and the network interface are both used to communicate with other devices. The processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores instructions that, when executed, perform the method as described in any one of claims 1 to 7.

Images