Laser-ultrasound-based melt inclusion detection system and method

Abstract

The application discloses a melt inclusion detection system and method based on laser ultrasound, and belongs to the technical field of ultrasonic detection. The method specifically comprises the following steps: a time-varying sound velocity compensation model of melt coupling of a container wall is established, a pulse laser is used to sequentially scan and excite at multiple annular discrete positions of a detection area, ultrasonic waves are generated in the container wall and coupled into the melt, a multi-propagation-direction detection field covering the interior of the melt is formed, time-domain echo signals of the annular discrete positions are collected, wall echo subtraction, flow background suppression and dispersion correction processing are performed on the time-domain echo signals, inclusion scattering signals are extracted, joint analysis is performed based on the multi-propagation-direction consistency and phase and spectral characteristics of the inclusion scattering signals, the existence and spatial position of inclusions in the melt are determined, and an inclusion detection result is output. According to the application, inclusion-related signals are extracted and integrated under multi-directional constraints, so that stable detection result output can be realized under complex melt working conditions.

Description

Technical Field

[0001] This invention belongs to the field of ultrasonic testing technology, specifically a melt inclusion detection system and method based on laser ultrasound. Background Technology

[0002] In metallurgical, continuous casting, and high-end alloy preparation processes, the presence of inclusions within the melt directly affects the purity of the material and the mechanical properties and reliability of subsequent products. Inclusions typically include oxide particles, slag inclusions, and bubbles, and their size distribution is wide and they exhibit dynamic changes during melt flow.

[0003] Traditional contact ultrasonic testing suffers from limitations in high-temperature molten environments, such as difficulty in maintaining the coupling medium, susceptibility to thermal damage to sensing elements, and significant signal attenuation, resulting in poor testing stability. Laser ultrasonic technology, due to its non-contact method of exciting and receiving ultrasonic signals, has the potential advantage of operating in high-temperature environments. However, in practical applications, it still faces challenges such as the complex propagation path of sound waves between the container wall and the melt, fluctuations in sound velocity with temperature and flow field variations, and the difficulty in accurately describing interface reflection and refraction behavior. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes a melt inclusion detection system and method based on laser ultrasound, which distinguishes echo differences caused by inclusions, bubbles, and flow disturbances in the melt environment, thus solving the problem of difficult inclusion identification and location.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] Laser-ultrasound-based methods for detecting inclusions in melts include:

[0007] A detection area is set on the outside of the container wall carrying the melt to obtain the container wall thickness, melt temperature parameters and geometric boundary parameters, and to establish a time-varying sound velocity compensation model for the coupling between the container wall and the melt.

[0008] By using pulsed laser to sequentially scan and excite multiple circumferential discrete positions in the detection area, ultrasonic waves are generated inside the container wall and coupled into the melt, forming a multi-propagation direction detection field covering the interior of the melt. The excitation parameters at each circumferential discrete position are adaptively adjusted according to the time-varying sound velocity compensation model.

[0009] Time-domain echo signals at discrete locations in each circumferential direction are acquired, and wall echo subtraction, flow background suppression, and dispersion correction are performed on the time-domain echo signals to extract inclusion scattering signals.

[0010] By combining the consistency of multiple propagation directions and the phase and spectral characteristics of inclusion scattering signals, the presence and spatial location of inclusions in the melt are determined, and the inclusion detection results are output.

[0011] Specifically, the establishment of the time-varying sound velocity compensation model for the container wall melt coupling includes:

[0012] The boundary outline of the detection area is set along the outer side of the container wall, and the detection area is divided into grid-like discrete parts to obtain multiple detection sub-regions arranged in directional order;

[0013] The wall penetration propagation path is extracted along the direction perpendicular to the local surface of the container wall, and the corresponding wall penetration path is segmented and calibrated according to the geometric boundary parameters of each detection sub-region to obtain the local wall thickness parameters and wall thickness continuity parameters of each detection sub-region.

[0014] The melt temperature parameters are mapped into multiple temperature action zones in a way that corresponds one-to-one with each detection sub-region. Combined with the local wall thickness parameters of each detection sub-region, the propagation time sequence association is marked for each temperature action zone, forming a time sequence correspondence between the detection sub-region and the temperature action zone.

[0015] Based on the local wall thickness parameters, wall thickness continuity parameters, geometric boundary parameters, and the time sequence correspondence of each detection sub-region, the segmented propagation sequence of laser ultrasound from the outside of the container wall into the inside of the container wall, coupled into the melt through the inside of the container wall, and continuing to propagate in the melt is determined in sequence, forming the interface propagation sequence corresponding to each detection sub-region.

[0016] Based on the propagation sequences of each interface, the segmented propagation order of different detection sub-regions is combined and organized to establish a time-varying sound velocity compensation model that includes wall thickness segments, temperature-affected segments, and boundary constraints of the detection area.

[0017] Specifically, the method of sequentially scanning and exciting multiple circumferential discrete positions of a pulsed laser in the detection area to generate and couple ultrasonic waves into the melt within the container wall, forming a multi-propagation direction detection field covering the interior of the melt, includes:

[0018] Using the geometric boundary of the detection area as a constraint, the detection area is divided into multiple excitation position units along the circumferential direction of the outer side of the container wall, and the corresponding circumferential excitation sequence is generated according to the preset orientation order;

[0019] Each excitation location unit is associated with the corresponding detection sub-region in the time-varying sound speed compensation model, and the propagation path constraint information corresponding to each excitation location unit is extracted to form a one-to-one path constraint set.

[0020] Based on the path constraint set, the incident position offset order and excitation timing interval of the pulse laser of each excitation position unit are set respectively, so that the adjacent excitation position units form a staggered excitation relationship and maintain the temporal distinguishability of each excitation path.

[0021] According to the circumferential excitation sequence, each excitation position unit is sequentially excited by pulsed laser. After completing one round of circumferential scanning, the incident position offset order is changed to form a new excitation sequence, so that multiple rounds of scanning form an interlaced ultrasonic field with different propagation directions inside the melt.

[0022] The staggered ultrasonic fields are summarized and rearranged according to their corresponding excitation position units and propagation path constraint information to construct a multi-propagation direction detection field covering the detection area.

[0023] Specifically, the step of summarizing and rearranging the staggered ultrasonic fields according to their corresponding excitation position units and propagation path constraint information to construct a multi-propagation direction detection field covering the detection area includes:

[0024] The ultrasonic propagation trajectories formed by each excitation position unit during each round of interleaved scanning are extracted into independent propagation segments, and then collected according to the excitation round, excitation position unit and corresponding path constraint information to form multiple sets of propagation segments to be rearranged.

[0025] Based on the orientation order and path constraint information of adjacent excitation position units in each set of propagation segments to be rearranged, the independent propagation segments are sorted by adjacency to determine the sequential splicing relationship and cross-switching relationship between different propagation segments, forming candidate propagation chains classified by propagation direction;

[0026] According to the incident bias order corresponding to the change of the excitation round, the propagation segments contained in each candidate propagation chain are replaced and rearranged across rounds, so that the propagation segments from different interleaved scanning rounds are merged under the same propagation direction category, resulting in multiple directional propagation chains.

[0027] The directional propagation chains are combined and arranged according to their coverage order within the detection area to establish a correspondence between propagation direction categories and coverage segments of the detection area, and a multi-propagation direction detection field covering the detection area is generated accordingly.

[0028] Specifically, the acquisition of time-domain echo signals at discrete locations in each circumferential direction, and the processing of these time-domain echo signals including wall echo subtraction, flow background suppression, and dispersion correction to extract inclusion scattering signals, includes:

[0029] According to the excitation sequence corresponding to each circumferential discrete position, the time-domain echo signals returned after each excitation are continuously acquired, and the time-domain echo signals are grouped and labeled according to the excitation position, excitation round and propagation direction category to form multiple sets of original echo sequences;

[0030] The wall main echo segments of the original echo sequence are extracted according to the time sequence under the same propagation direction category. The wall main echo segments in the same circumferential discrete position in different excitation rounds are aligned and sorted to form the reference wall echo sequences corresponding to each circumferential discrete position.

[0031] Using the reference wall echo sequence corresponding to each circumferential discrete position as a reference, the corresponding echo segments in the original echo sequence of the same group are sequentially subtracted by segment replacement and spliced ​​with neighboring residuals to obtain the intermediate echo sequence after removing the main wall echo.

[0032] For the intermediate echo sequence, echo segments with close time are extracted according to the circumferential correspondence between adjacent excitation cycles, and cross-cycle differential rearrangement and same-direction segment merging are performed on the extracted echo segments to form an echo sequence to be corrected after removing the flow background.

[0033] Based on the time-varying sound speed compensation model, the echo sequence to be corrected is segmented and corrected according to the corresponding propagation direction category and path constraint information. Time delay anomalous segments, phase anomalous segments, and energy mutation segments are extracted from each corrected echo segment. The time delay anomalous segments, phase anomalous segments, and energy mutation segments are correlated and aggregated to obtain the inclusion scattering signal.

[0034] Specifically, based on the time-varying sound speed compensation model, the echo sequence to be corrected is segmented for propagation correction according to the corresponding propagation direction category and path constraint information. Time delay anomaly segments, phase anomaly segments, and energy mutation segments are extracted from each corrected echo segment. These time delay anomaly segments, phase anomaly segments, and energy mutation segments are correlated and aggregated to obtain the inclusion scattering signal, including:

[0035] The echo sequences to be corrected are classified and expanded according to their corresponding propagation direction categories. Combined with the path constraint information corresponding to each echo sequence, each echo sequence to be corrected is sequentially mapped into an in-wall propagation segment, an interface transition segment, and a melt propagation segment, forming a correction segment sequence corresponding to each propagation direction category.

[0036] Based on the time-varying sound velocity compensation model, the wall propagation segment, interface transition segment and melt propagation segment in each of the correction segment sequences are aligned and rearranged in order, so that the corresponding segments from different path constraint information under the same propagation direction category form a segmented correction echo chain according to the propagation sequence.

[0037] Echo segments whose arrival order has shifted, echo segments whose connection with adjacent segments has broken down, and echo segments that exhibit isolated transitions in the same-direction segmented correction echo chains are extracted sequentially along the propagation direction and labeled as time delay anomalous segments, phase anomalous segments, and energy mutation segments, respectively.

[0038] The time delay anomaly segments, phase anomaly segments, and energy mutation segments extracted under the same propagation direction category are associated and arranged according to the occurrence segment, adjacent order, and cross-direction overlap relationship. The segments that meet the preset association order are combined into candidate scattering units. Then, the candidate scattering units in different propagation direction categories are merged accordingly to obtain the inclusion scattering signal.

[0039] Specifically, the joint analysis based on the consistency of multiple propagation directions and the phase and spectral characteristics of inclusion scattering signals determines the presence and spatial location of inclusions in the melt, and outputs inclusion detection results, including:

[0040] The inclusion scattering signals are grouped and organized according to the corresponding propagation direction category, excitation position unit and path constraint information to form multiple scattering unit sequences corresponding to different propagation directions, and the occurrence order information in each scattering unit sequence is recorded.

[0041] Based on the occurrence order information of each scattering unit sequence, the scattering units in different propagation directions are cross-compared. The scattering units that satisfy the temporal connection relationship between adjacent propagation directions are grouped into the same source candidate group, and the scattering units that do not form the same source candidate group are retained as independent candidate units.

[0042] Phase shift segments, spectral shift segments, and frequency band split segments are extracted from the scattering signals corresponding to the same candidate group and independent candidate unit, respectively. The phase shift segments, spectral shift segments, and frequency band split segments are jointly indexed according to the propagation direction category to form phase spectral feature sequences corresponding to each candidate target.

[0043] Based on the phase spectrum feature sequence, and combined with the propagation direction category, excitation position unit and path constraint information, the intersection position of each candidate target in the detection area is defined, and the spatial position of the inclusion target in the melt is determined.

[0044] The determined spatial locations of inclusions are archived, and inclusion detection results containing inclusion presence identifiers, their respective detection areas, and corresponding spatial locations are generated and output.

[0045] Specifically, the step of defining the intersection position of each candidate target within the detection area based on the phase spectrum feature sequence, combined with the propagation direction category, excitation position unit, and path constraint information, to determine the spatial position of the inclusion target in the melt includes:

[0046] The phase spectrum feature sequences corresponding to each candidate target are expanded according to the propagation direction category, and the directional action segments corresponding to each candidate target under each propagation direction are determined based on the excitation position unit and path constraint information corresponding to each propagation direction category, forming candidate positioning segment sequences corresponding to different propagation directions respectively.

[0047] For each candidate positioning segment sequence, according to the arrangement order of the corresponding excitation position units, the action segments of each direction are sequentially delivered to the corresponding crossing segments inside the detection area, and the connecting boundaries of the action segments of each direction are sequentially adjusted according to the path constraint information to form a set of multiple direction delivery segments.

[0048] The delivery segment set is cross-matched pairwise. Delivery segments with overlapping and crossing relationships within the detection area and whose phase spectrum feature sequences maintain the corresponding order are merged into the same candidate intersection range. Each candidate intersection range is then converged step by step according to the number of propagation directions and the order of overlap to form the target positioning area.

[0049] The distribution order of the target positioning areas and the corresponding candidate target excitation position units is correlated and verified. The positions of the target positioning areas that satisfy the same distribution order constraint are determined, and the determined positions are determined as the spatial positions of the inclusion targets in the melt.

[0050] Specifically, the process of archiving the determined spatial locations of inclusions, generating and outputting inclusion detection results that include inclusion presence identifiers, their respective detection areas, and corresponding spatial locations, includes:

[0051] The spatial positions of each inclusion target in the melt are organized according to the distribution order within the detection area, and position unit records are established for each inclusion target based on the propagation direction category, excitation position unit and path constraint information of the corresponding candidate target.

[0052] Each location unit record is linked to the detection area boundary and its internal segment division to determine the detection area to which each inclusion target belongs and the position order within the area, and an area affiliation mark is written into the corresponding location unit record.

[0053] For each location unit record with the write area belonging to the mark, the fields are arranged in the order of the inclusion presence identifier, the detection area to which it belongs and the corresponding spatial location. The location unit records in the same detection area are merged in the order of location to form the corresponding inclusion detection result entries.

[0054] The laser-ultrasound-based melt inclusion detection system is used to implement the laser-ultrasound-based melt inclusion detection method, and includes: a parameter acquisition module, a detection field construction module, a signal extraction module, and an inclusion detection module.

[0055] The parameter acquisition module is used to set a detection area on the outside of the container wall carrying the melt, acquire the container wall thickness, melt temperature parameters and geometric boundary parameters, and establish a time-varying sound velocity compensation model for the coupling of the container wall and the melt.

[0056] The detection field construction module is used to use pulsed laser to sequentially scan and excite multiple circumferential discrete positions in the detection area, generate ultrasonic waves in the container wall and couple them into the melt, forming a multi-propagation direction detection field covering the interior of the melt.

[0057] The signal extraction module is used to collect time-domain echo signals at discrete locations in each circumferential direction, and to perform wall echo subtraction, flow background suppression and dispersion correction on the time-domain echo signals to extract inclusion scattering signals.

[0058] The impurity detection module is used to perform joint analysis based on the consistency of multiple propagation directions and the phase and spectral characteristics of the impurity scattering signal, to determine the presence and spatial location of impurities in the melt, and to output the impurity detection results.

[0059] Compared with the prior art, the beneficial effects of the present invention are:

[0060] This invention proposes a laser-ultrasound-based melt inclusion detection system and method. It generates a multi-propagation-direction ultrasonic detection field through staggered laser excitation at multiple circumferential positions. The acquired echo signals are sequentially processed with wall echo subtraction, flow background suppression, and segmented propagation correction. Based on this, joint analysis is performed using the consistency of multiple propagation directions and phase and spectral characteristics to achieve the identification and spatial positioning of inclusions within the melt. This invention can complete the detection process without contacting the high-temperature melt and effectively distinguishes echo signals in complex propagation environments through multi-path information reconstruction and hierarchical signal processing. This allows inclusion-related signals to be extracted and integrated under multi-directional constraints, thereby achieving stable detection result output under complex melt conditions. Attached Figure Description

[0061] Figure 1 A flowchart of the melt inclusion detection method based on laser ultrasound provided by the present invention;

[0062] Figure 2 The scattering signal and abnormal segment diagram provided by this invention;

[0063] Figure 3 This is a schematic diagram of the spatial positioning of inclusions provided by the present invention;

[0064] Figure 4 The diagram shows the architecture of the laser-ultrasound-based melt inclusion detection system provided by this invention. Detailed Implementation

[0065] The present application will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present application, but do not limit the present application in any way. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present application. These all fall within the protection scope of the present application.

[0066] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0067] It should be noted that, unless there is a conflict, the various features in the embodiments of this application can be combined with each other, all of which are within the protection scope of this application. Furthermore, although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than the module division in the device or the order in the flowchart. In addition, the terms "first," "second," and "third" used in this application do not limit the data or execution order, but only distinguish identical or similar items with essentially the same function and effect.

[0068] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The term "and / or" as used in this specification includes any and all combinations of one or more of the associated listed items.

[0069] Example 1

[0070] Please see Figure 1 The present invention provides an embodiment of a method for detecting inclusions in melts based on laser ultrasound, comprising the following specific steps:

[0071] Step S1: For the outer side of the container wall carrying the melt, a detection area is preset. The container wall thickness distribution, melt temperature parameters and detection area geometric boundary parameters corresponding to the detection area are obtained. Based on the wall thickness distribution, melt temperature parameters and detection area geometric boundary parameters, a time-varying sound velocity compensation model including the container wall-melt coupling interface is established to characterize the propagation delay, refraction path and dispersion shift of laser ultrasound in the container wall and melt.

[0072] The specific steps for establishing a time-varying sound velocity compensation model for the coupling between the container wall and the melt are as follows:

[0073] First, using the outer wall of the container as the reference plane, for the target section of the container wall to be inspected, existing surface unfolding and two-dimensional projection calibration methods are used to project the cylindrical or irregular curved surface of the outer side of the container wall onto a two-dimensional plane along its normal. A closed detection area boundary contour is then defined along the circumferential ends and axial upper and lower ends corresponding to the target section within the projection plane. This boundary contour completely encloses the outer region of the container wall to be inspected, and the geometric dimensions of the boundary contour are mapped to the actual dimensions of the target section of the container wall at a 1:1 ratio, without scaling or distortion. Subsequently, existing structured mesh generation methods are used... The method involves dividing the detection area enclosed by the boundary contour into a grid. Specifically, based on the preset detection accuracy and detection element layout resolution, the circumferential step size and axial step size of the grid unit are set. The circumferential step size is no greater than the arc length corresponding to a 5° central angle in the circumferential direction of the container, and the axial step size is no greater than 100mm. The detection area is fully covered by an equidistant structured quadrilateral grid. During the division process, the boundaries of adjacent grid units are controlled to be seamlessly connected, without overlap or gaps. Each independent grid unit obtained after the discrete division is a detection sub-region.

[0074] Finally, according to the clockwise orientation of the container's circumference and the top-to-bottom orientation of the axial direction, all detection sub-regions are sequentially and uniquely numbered, resulting in multiple detection sub-regions arranged in spatial orientation order. Each detection sub-region corresponds to a unique spatial location and fixed geometric area on the outside of the container wall, and all detection sub-regions completely cover the entire detection area.

[0075] For each detection sub-region that has been divided and ordered, the unit normal vector of the local surface of the container wall corresponding to a single detection sub-region is first calculated using the three-dimensional surface micro-element normal method. The direction of this unit normal vector is the direction of wall-penetrating propagation perpendicular to the local surface of the container wall. Then, ultrasonic pulse echo detection is used. On the outer surface of the container wall covered by the detection sub-region, several detection sampling points are set up along the circumferential and axial directions at preset equidistant sampling steps. The central axis of the ultrasonic probe at each sampling point is completely coincident with the unit normal vector corresponding to that sampling point. In this way, the complete wall-penetrating propagation path along the normal direction through the container wall at each sampling point is extracted. This path is the one-way effective propagation path of the ultrasonic pulse, which is perpendicularly incident from the outer incident surface of the container wall, and returns along the original path through the inner reflecting surface of the container wall. Then, based on the pre-generated geometric boundary parameters of the detection sub-region, including the spatial coordinates of the circumferential start and end boundaries of the sub-region, the radius of curvature of the surface covered by the sub-region, and the effective coverage area, all wall-penetrating propagation paths in the sub-region are segmented and calibrated.

[0076] Specifically, through spatial coordinate matching and filtering, valid wall-penetrating paths whose incident points fall entirely within the geometric boundary of the detection sub-region are retained, while invalid paths whose incident points cross the sub-region boundary are eliminated. At the same time, the complete propagation interval of each valid wall-penetrating path in the direction of the container wall thickness is calibrated, thus completing the segmented calibration of the wall-penetrating path corresponding to a single detection sub-region.

[0077] After calibration, the existing ultrasonic time-of-flight method is used to collect the time-of-flight difference between the ultrasonic incident wave and the reflected echo from the inner side of the container wall corresponding to each effective wall penetration path. Combined with the ultrasonic longitudinal wave propagation velocity of the container wall substrate calibrated in advance using a standard test block of the same material, the measured value of the single-point wall thickness corresponding to each effective wall penetration path is calculated. The arithmetic mean of all the measured values ​​of the single-point wall thickness in a single detection sub-region is calculated to obtain the local wall thickness parameter of the detection sub-region. The maximum value, minimum value and standard deviation of the measured values ​​of the single-point wall thickness in the sub-region are recorded simultaneously.

[0078] Finally, based on the orientational arrangement of all detection sub-regions, the local wall thickness parameters of adjacent detection sub-regions and the measured wall thickness values ​​at single points on adjacent boundaries are calculated using first-order difference to obtain the wall thickness gradient values ​​between adjacent sub-regions. Simultaneously, the variance of all measured wall thickness values ​​at single points within a single detection sub-region is calculated to obtain the wall thickness uniformity parameter within the sub-region. The aforementioned wall thickness gradient values, wall thickness uniformity parameters, and wall thickness abrupt change results between adjacent sub-regions determined based on a preset wall thickness abrupt change threshold are used together as the wall thickness continuity parameters corresponding to each detection sub-region.

[0079] After acquiring the local and continuous wall thickness parameters of each detection sub-region, real-time melt temperature parameters distributed along the axial and circumferential directions of the melt inside the container are obtained through distributed multi-point temperature measurement. The number of acquisition points for these melt temperature parameters is exactly the same as the number of detection sub-regions, and the spatial position of each acquisition point corresponds one-to-one with the geometric center of the corresponding detection sub-region in the radial direction of the container. The melt temperature parameters acquired at each acquisition point are mapped to temperature action segments bound to the corresponding detection sub-regions according to spatial matching. The circumferential and axial boundary ranges of each temperature action segment correspond to the geometric boundaries of the corresponding detection sub-regions. The parameters are completely overlapped. At the same time, combined with the local wall thickness parameters of each detection sub-region, the propagation time of the ultrasonic pulse in the container wall of the corresponding detection sub-region and the propagation time in the melt of the corresponding temperature action zone are calculated using the existing ultrasonic propagation time sequence calculation method. Taking the moment when the ultrasonic pulse is incident from the outside of the container wall as the time sequence zero point, the start and end times of ultrasonic propagation in the melt corresponding to each temperature action zone are marked to complete the propagation time sequence association marking of each temperature action zone. Finally, a time sequence correspondence relationship is formed, which binds the spatial position of the detection sub-region, the local wall thickness parameters, the temperature value of the temperature action zone, and the propagation time sequence interval.

[0080] Based on the local wall thickness parameters, wall thickness continuity parameters, geometric boundary parameters generated for each detection sub-region, and the temporal correspondence between the detection sub-region and the temperature-affected zone, the unit normal vector corresponding to a single detection sub-region is used as the propagation axis. According to the order of ultrasonic propagation, the three-segment propagation sequence of the laser ultrasound is determined sequentially. The first segment is the propagation segment inside the container wall where the laser ultrasound enters the container wall from the outer incident surface and propagates along the propagation axis towards the inner side of the container wall. The propagation distance of this segment is the local wall thickness parameter of the corresponding detection sub-region. The propagation time interval is calculated and determined by the sound velocity of the container wall substrate and the local wall thickness parameters. The propagation boundary constraints are determined by the geometric boundary parameters and wall thickness continuity parameters of the corresponding detection sub-region.

[0081] The second segment is the transmission propagation segment of laser ultrasound through the solid-liquid coupling interface on the inner side of the container wall, coupling from the container wall substrate into the internal melt. The transmission timing node of this segment completely coincides with the propagation termination timing node of the first segment, and the transmission timing node is time-matched with the propagation start timing node of the corresponding temperature action zone.

[0082] The third segment is the internal propagation segment of the melt, which continues to propagate along the propagation axis after the laser ultrasound completes the solid-liquid interface transmission. The propagation time interval of this segment is completely matched with the propagation time interval of the corresponding temperature action segment, and the propagation speed is constrained by the melt temperature parameter of the corresponding temperature action segment. The above three-segment propagation sequence is sorted according to the propagation time sequence to form an interface propagation sequence that uniquely corresponds to the detection sub-region. All detection sub-regions generate their respective corresponding interface propagation sequences in the above manner.

[0083] Finally, based on the interface propagation sequence corresponding to each detection sub-region, and according to the orientational arrangement of the detection sub-regions, the segmented propagation sequences corresponding to different detection sub-regions are combined and organized in the same dimension. The container wall propagation segments corresponding to all detection sub-regions are integrated into the model's wall thickness segment module. This wall thickness segment module embeds the local wall thickness parameters, wall thickness continuity parameters, and geometric boundary parameters of each detection sub-region. All temperature action segments corresponding one-to-one with the detection sub-regions are integrated into the model's temperature action segment module. This temperature action segment module embeds the real-time melt temperature parameters of each temperature action segment and the temporal correspondence with the corresponding detection sub-region. The boundary contour of the detection area is integrated into the model's detection area boundary constraint module, which is used to limit the effective range of the model and the spatial matching relationship of each module.

[0084] Finally, based on the above three modules, and combined with the segmented propagation time sequence and sound velocity-temperature correlation characteristics corresponding to the propagation sequence of each interface, a time-varying sound velocity compensation model is established, which includes the wall thickness section, the temperature effect section, and the boundary constraints of the detection area. This model can output the real-time compensation value of the ultrasonic propagation sound velocity at the corresponding detection sub-region position based on the input real-time melt temperature parameters, and complete the dynamic compensation of sound velocity under the coupling state of the container wall melt.

[0085] It should be noted that the construction of the time-varying sound velocity compensation model is not only based on geometric path division, but also includes the construction of a quantitative mapping relationship between sound velocity and temperature and the calculation of segmented propagation time. Specifically, the sound velocity variation curve of the melt material under different temperature conditions is first obtained through pre-experiment calibration, and the curve is discretized and stored as a temperature-sound velocity mapping table. In the actual detection process, the corresponding melt propagation sound velocity is determined by looking up the table or interpolation calculation based on the real-time temperature parameters of each temperature zone.

[0086] Step S2: Control the pulsed laser beam to act on the first wall position of the detection area, and excite broadband ultrasonic waves coupled into the melt along the wall thickness direction inside the container wall. According to the preset timing sequence, the pulsed laser beam is sequentially scanned and excited at multiple circumferential discrete positions around the first wall position to form a multi-propagation direction ultrasonic detection field that crosses and covers inside the melt. The excitation parameters of each circumferential discrete position are adaptively adjusted according to the time-varying sound velocity compensation model so that the ultrasonic echoes of different propagation directions are comparable and aligned under the same time base.

[0087] The specific steps for forming a multi-propagation direction detection field covering the interior of the melt are as follows:

[0088] Using the closed geometric boundary of the detection area in the time-varying sound velocity compensation model as a spatial constraint, and employing the existing cylindrical surface circumferential equidistant division method, the detection area is divided into multiple excitation position units arranged sequentially along the circumferential direction of the container wall, with a preset circumferential step size not exceeding the arc length corresponding to a central angle of 3° in the circumferential direction of the container. This results in multiple excitation position units that are arranged sequentially along the circumferential direction and whose axial coverage is completely consistent with the axial boundary of the detection area. The circumferential boundary of each excitation position unit has no overlap, no gaps, and completely covers the entire circumferential range of the detection area. Subsequently, according to the preset clockwise orientation order of the container circumferential direction, all excitation position units are continuously and uniquely numbered to generate a circumferential excitation sequence that corresponds one-to-one with the numbering sequence, thus clarifying the excitation order of each excitation position unit during a single-round circumferential scan.

[0089] After generating the circumferential excitation sequence, the spatial coordinate range of each excitation location unit is spatially matched and associated with the corresponding detection sub-region in the time-varying sound velocity compensation model. Specifically, all detection sub-regions within the coverage area of ​​a single excitation location unit are bound to that excitation location unit. Subsequently, based on the time-varying sound velocity compensation model, the geometric boundary parameters, local wall thickness parameters, wall thickness continuity parameters, interface propagation sequence, and time sequence correspondence of all detection sub-regions bound to each excitation location unit are extracted. These parameters are integrated into propagation path constraint information corresponding to that excitation location unit, which clearly defines the effective propagation direction, propagation boundary, solid-liquid coupling transmission constraint, and time sequence matching rules of the pulsed laser-excited ultrasonic wave. Finally, the propagation path constraint information corresponding to all excitation location units is collected according to the numbering order of the circumferential excitation sequence to form a path constraint set that corresponds one-to-one with the excitation location unit.

[0090] It should be noted that the path constraint information is used to limit the propagation behavior of ultrasonic waves in the container wall and melt. Specifically, it includes: the coordinates of the propagation start position, the propagation direction vector, the propagation path length, the interface position, and the corresponding propagation time interval. Among them, the propagation direction vector is determined by the normal vector corresponding to the detection sub-region, the propagation path length is determined by the local wall thickness parameter and the melt propagation depth, the interface position is determined by the geometric boundaries of the inner and outer surfaces of the container wall, and the propagation time interval is calculated by the time-varying sound velocity compensation model.

[0091] Based on the generated set of path constraints, for each excitation position unit, firstly, taking the circumferential center line of the excitation position unit as a reference, multiple equally spaced laser incident points are set sequentially from top to bottom along the axial direction. The axial spacing between adjacent incident points is consistent with the axial step size of the detection sub-region, thereby setting the incident position offset order of the pulsed laser in the unit.

[0092] The feasible incident position offset sequence is divided into two categories: axial forward offset (from top to bottom) and axial reverse offset (from bottom to top). At the same time, based on the ultrasonic wave propagation time interval corresponding to adjacent excitation position units in the path constraint set, the excitation time interval between adjacent excitation position units is set. This time interval is greater than the maximum propagation time of the ultrasonic wave excited by a single excitation position unit in the container wall and melt, ensuring that the ultrasonic waves generated by adjacent excitations do not overlap in the time domain. At the same time, the incident position offset sequence of adjacent excitation position units is reversed, forming a staggered excitation relationship between circumferential adjacent units, thereby maintaining the uniqueness and distinguishability of the excitation path corresponding to each excitation position unit in the time sequence.

[0093] According to the preset circumferential excitation sequence, existing nanosecond pulsed laser excitation is used to sequentially excite each excitation position unit according to the set incident position offset order and excitation time interval. The feasible pulsed laser parameters are wavelength 1064nm, pulse width 5-20ns, and single pulse energy 10-50mJ. Broadband ultrasonic waves are excited inside the container wall through the thermoelastic effect. The ultrasonic waves propagate through the wall in a direction perpendicular to the local surface of the container wall, as determined by the corresponding propagation path constraint information. They are transmitted into the melt through the solid-liquid coupling interface inside the container wall. After completing a single round of circumferential scanning excitation, the incident position offset order of all excitation position units is changed, and the starting excitation position unit of the circumferential excitation sequence is adjusted to form a new circumferential excitation sequence. The next round of staggered scanning excitation is completed according to the new excitation sequence. The above multi-round scanning process is repeated. Each round of scanning corresponds to a different incident position offset order and starting excitation position, so that the ultrasonic waves excited by the multi-round scanning form a staggered ultrasonic field with different propagation directions and spatial trajectories inside the melt.

[0094] After completing multiple rounds of staggered scanning excitation, the ultrasonic propagation trajectories formed by pulsed laser excitation at each excitation position unit during each round of staggered scanning, along with the corresponding path constraint information, are first extracted into independent propagation segments carrying the corresponding excitation round number, excitation position unit number, propagation path constraint information, and time domain identification information. Then, according to the excitation round and excitation position unit numbering order, all independent propagation segments are grouped together to form multiple sets of propagation segments to be rearranged, each corresponding to all propagation segments of a single excitation round. Subsequently, based on the azimuth order of adjacent excitation position units and their corresponding path constraint information in each set of propagation segments to be rearranged, the independent propagation segments within the set are sorted by adjacency.

[0095] Specifically, the spatial angle between the propagation directions is no greater than 5° as the threshold for determining the same direction, thus determining the sequential splicing relationship between different propagation segments. The crossover switching criteria are determined by the existence of spatial intersections in the propagation trajectories and the spatial angle between the propagation directions being greater than 30°, thus determining the crossover switching relationship between different propagation segments. This ultimately forms multiple candidate propagation chains classified according to the ultrasound propagation direction. Each candidate propagation chain corresponds to a propagation direction category and includes all adjacent independent propagation segments in the same direction. Subsequently, according to the incident offset order corresponding to the change in excitation rounds, the independent propagation segments contained in each candidate propagation chain are replaced and rearranged across rounds. Independent propagation segments in different excitation rounds and in the same propagation direction category are merged into the same candidate propagation chain. Duplicate propagation trajectory segments are removed, and trajectory gaps in the propagation direction are filled, so that propagation segments from different interleaved scanning rounds are merged in the same propagation direction category, resulting in multiple directional propagation chains corresponding to different propagation directions.

[0096] Finally, all directional propagation chains are combined and arranged according to their coverage order in the corresponding container space of the detection area, establishing a one-to-one correspondence between each propagation direction category and the corresponding coverage segment inside the detection area. At the same time, the propagation coverage of all directional propagation chains is integrated to ensure that the integrated coverage completely covers the entire space of the melt inside the container corresponding to the detection area, thereby generating a multi-propagation direction detection field that covers the melt inside the detection area.

[0097] Step S3: Using a laser vibration measurement unit located at a second wall position separate from the first wall position, synchronously acquire the emission response of the multi-propagation direction ultrasonic detection field on the container wall to obtain time-domain echo signals corresponding to each discrete circumferential position; sequentially perform wall intrinsic echo subtraction, flow background differential elimination based on adjacent time sequences, and dispersion correction processing based on the time-varying sound velocity compensation model on the time-domain echo signals to extract the scattering enhancement component caused by discrete inclusions inside the melt, and suppress the signals caused by melt surface disturbance, transient bubble rupture, and overall flow.

[0098] The specific steps for extracting the scattering signal from inclusions are as follows:

[0099] Using an existing laser vibrometer (laser vibration unit), the center wavelength phase anomaly segment is 532nm, the velocity resolution is ≤1μm / s, the sampling frequency is ≥100MHz, and the frequency response range covers the phase anomaly segment from 0.5MHz to 10MHz. Matching the ultrasonic main frequency range of pulsed laser excitation, according to the excitation sequence corresponding to each circumferential discrete position, the time-domain echo signals returning along the original propagation path after each pulsed laser excitation are continuously and synchronously acquired. The duration of a single acquisition is not less than twice the maximum propagation time of a single-excitation ultrasonic wave in the container wall and melt, ensuring complete coverage of the echo propagation period of the entire depth of the melt. Subsequently, based on the excitation position unit number, excitation cycle number, and propagation direction category corresponding to each time-domain echo signal, all acquired time-domain echo signals are uniquely grouped and labeled. Each group of echo signals corresponds to a unique excitation position, excitation cycle, and propagation direction category, forming multiple sets of original echo sequences carrying complete spatial, temporal, and directional identifiers.

[0100] After constructing the original echo sequence, all original echo sequences under the same propagation direction category are sorted according to the order of excitation time. Based on the local wall thickness parameters of the corresponding detection sub-region and combined with the pre-calibrated ultrasonic longitudinal wave velocity of the container wall substrate, the theoretical arrival time window of the main wall echo is calculated. In this way, the main wall echo segments that fall within the time window and are reflected from the inner and outer surfaces of the container wall in each set of original echo sequences are accurately extracted. Then, the main wall echo segments in different excitation rounds, located at the same circumferential discrete position and with the same propagation direction category, are time-domain aligned and sorted with the peak time of the incident wave on the wall as the reference. The arithmetic mean filter is applied to the aligned sets of main wall echo segments to eliminate random noise, and finally, the reference wall echo sequences corresponding to each circumferential discrete position and each propagation direction category are formed.

[0101] Using the reference wall echo sequences corresponding to each discrete position in the circumferential direction and each propagation direction category as a reference, the corresponding echo segments in the original echo sequences that are time-domain aligned with the reference wall echo sequences are sequentially replaced by segment substitution. Specifically, the time-domain data of the main wall echo segments in the original echo sequences are directly replaced with zero-amplitude sequences of equal duration. At the same time, residual smoothing is performed on the neighboring data of the five sampling points before and after the replacement segment to eliminate the step noise caused by the segment replacement. In this way, the strong amplitude main wall echo components in the original echo sequences are completely removed, resulting in the intermediate echo sequence after removing the main wall echo.

[0102] For the obtained intermediate echo sequence, according to the circumferential correspondence between adjacent excitation rounds, echo segments of equal duration and time proximity in the same circumferential discrete position, under the same propagation direction category, and corresponding melt propagation segment are extracted from adjacent excitation rounds. Cross-round differential rearrangement is performed on the extracted multiple sets of echo segments. That is, by performing time-domain differential operation on the corresponding echo segments of adjacent rounds, the steady-state flow background noise caused by melt flow is eliminated. Then, the echo segments after differential processing are merged into segments in the same direction according to the propagation direction category. The echo segments under the same propagation direction category, all circumferential discrete positions, and all excitation rounds are rearranged in time sequence to form the echo sequence to be corrected after removing the flow background.

[0103] Based on the time-varying sound velocity compensation model, the echo sequence to be corrected is first classified and expanded according to its corresponding propagation direction category. Combined with the path constraint information corresponding to each echo sequence, each echo sequence to be corrected is sequentially mapped and divided into the wall propagation segment, the interface transition segment, and the melt propagation segment according to the time sequence. The start and end times of each segment in the time domain are precisely determined by the propagation time sequence interval in the path constraint information, forming a correction segment sequence corresponding to each propagation direction category.

[0104] Subsequently, based on the time-varying sound velocity compensation model, the wall propagation segment, interface transition segment, and melt propagation segment in each correction segment sequence were aligned in the time domain and rearranged in the segment. Specifically, based on the theoretical propagation time of the corresponding segment output by the time-varying sound velocity compensation model, the actual acquired segment time domain length was linearly stretched or compressed to eliminate the dispersion and time shift errors caused by ultrasonic speed fluctuations due to the time-varying melt temperature. At the same time, corresponding segments from different path constraint information under the same propagation direction category were orderly spliced ​​according to the order of ultrasonic wave propagation to form a segmented correction echo chain without time domain misalignment and dispersion error.

[0105] like Figure 2 As shown, along the direction of ultrasonic wave propagation, abnormal segments in each segment of the corrected echo train are extracted sequentially. Specifically, as shown... Figure 2 As shown above, in the first type of signal, by comparing the arrival time of the overall waveform of the echo signal, local segments with significant time shifts relative to the preceding and following signals are identified, and these local segments are marked as time delay aberration segments; such as Figure 2 As shown in the middle section, in the second type of signal, by analyzing the periodic changes of the echo signal, local segments that are misaligned in periodic structure with adjacent waveforms are identified, and these local segments are marked as phase anomalous segments; as shown in the middle section. Figure 2 As shown in the lower part, in the third type of signal, by detecting the change in the amplitude of the echo signal, the segment where the signal amplitude changes abruptly in a local area is identified, and the segment is marked as an energy change segment.

[0106] In this embodiment, the determination of time delay anomaly segments, phase anomaly segments, and energy mutation segments are all based on preset determination rules. Specifically, time delay anomaly segments are determined by comparing the difference between the actual arrival time of the echo signal and the theoretical arrival time calculated based on the time-varying sound speed compensation model. When the difference exceeds a preset time threshold, it is determined to be a time delay anomaly segment. Phase anomaly segments are determined by calculating the phase difference change of the echo signal in adjacent periods. When the phase offset exceeds a preset phase threshold, it is marked. Energy mutation segments are determined by analyzing the rate of change of the echo signal amplitude. When the rate of change of the amplitude exceeds a preset energy threshold, it is determined. The threshold parameters can be preset or adaptively adjusted according to the type of detection object, melt state, and equipment calibration results.

[0107] Finally, the time delay anomaly segments, phase anomaly segments, and energy mutation segments extracted under the same propagation direction category are associated and arranged according to their propagation segment, temporal adjacent order, and spatial overlap across propagation directions. Anomaly segments appearing in the same spatial location under different propagation direction categories are matched and associated. Segments that meet the preset association order, i.e., at least two of the three characteristics of time delay anomaly, phase decoupling, and energy mutation appearing at the same spatial location, are combined into candidate scattering units. Then, candidate scattering units with consistent spatial locations in different propagation direction categories are merged accordingly to eliminate artifact interference caused by single-direction detection. Finally, the inclusion scattering signal of the inclusion inside the melt, carrying accurate spatial location and temporal feature information, is obtained.

[0108] Step S4: Perform joint inversion of the scattering enhancement component according to the consistency of propagation direction, echo phase lag characteristics, spectral centroid drift characteristics, and multi-angle scattering energy focusing characteristics to determine the spatial location, equivalent size, and existence probability of suspected inclusions inside the melt. When the same spatial location meets the preset phase lag threshold, spectral centroid shift threshold, and energy focusing threshold in at least two different propagation directions, it is determined that there is an inclusion at that spatial location, and the inclusion detection result is output.

[0109] The specific steps for outputting the inclusion detection results are as follows:

[0110] The extracted inclusion scattering signals are classified and organized according to their corresponding propagation direction category, excitation position unit number and path constraint information. Scattering signals from the same excitation position unit and with the same propagation path constraint under the same propagation direction category are arranged in sequence to form multiple scattering unit sequences corresponding to different propagation directions. The temporal order of each scattering unit in the sequence, the peak amplitude time and the spatial mapping number are recorded simultaneously.

[0111] Based on the occurrence order information of each scattering unit sequence, the scattering units under different propagation directions are cross-compared in both spatial and temporal dimensions. The criterion is that the temporal difference is less than 3 sampling periods and the spatial projection angle is less than 10°. The scattering units that satisfy the temporal connection relationship between adjacent propagation directions and whose spatial orientation tends to be consistent are grouped into the same source candidate group. For the scattering units that only appear in a single propagation direction and cannot form temporal and spatial matching with scattering units in other directions, they are directly retained as independent candidate units.

[0112] Feature segments are extracted from the scattering signals of each scattering unit within the same candidate group and the independent candidate unit. Specifically, these include phase transfer segments where the phase changes abruptly compared to the theoretically straight propagation path, spectral shift segments where the peak frequency shifts compared to the melt background frequency domain reference, and frequency band splitting segments where a single frequency band splits into multiple discrete sub-frequency bands. The above three types of feature segments are uniformly indexed and sorted according to the propagation direction category to form phase spectral feature sequences corresponding to each same candidate group and independent candidate unit.

[0113] The phase spectrum feature sequences corresponding to each candidate target are expanded according to the propagation direction category. Based on the excitation position unit coordinates and the propagation axis and effective propagation interval in the path constraint information corresponding to each propagation direction category, the directional action segment of each candidate target under the corresponding propagation direction is determined. This interval is the theoretical crossing interval of the ultrasonic wave from the excitation position to the scattering point, thus forming a candidate positioning segment sequence corresponding to different propagation directions.

[0114] For each candidate positioning segment sequence, according to the circumferential arrangement of the excitation position units, the action segments of each direction are projected into the three-dimensional space of the melt corresponding to the detection area through spatial coordinate transformation to obtain the corresponding crossing segments. Based on the propagation time sequence and wall thickness boundary in the path constraint information, the connection boundary between the front and rear of each action segment is sequentially adjusted to remove invalid segments that exceed the effective propagation range, forming multiple sets of directional projection segments containing spatial crossing intervals and feature information.

[0115] By performing pairwise cross-comparisons on all delivery segments in all directions, delivery segments that overlap and cross each other in the three-dimensional space of the melt and whose corresponding phase spectrum feature sequences maintain the same order are merged into the same candidate intersection range. Then, each candidate intersection range is converged step by step according to the number of propagation directions covered and the degree of overlap. The more propagation directions overlap, the higher the convergence weight, and finally, a target positioning area with converged accuracy is formed.

[0116] The circumferential distribution order of each target positioning area and the corresponding candidate target excitation position unit is correlated and verified. The target positioning area that meets the same excitation distribution order constraint and has a spatial overlap degree higher than the preset threshold is positioned. The geometric center of the overlapping area is taken as the final coordinate, and the positioned position is determined as the spatial position of the inclusion target in the melt.

[0117] Subsequently, the spatial positions of each inclusion target are organized according to the circumferential and axial distribution order within the detection area. Based on the propagation direction category, excitation position unit, and path constraint information of the corresponding candidate targets, position unit records containing coordinate parameters, feature identifiers, and excitation sources are established for the spatial positions of each inclusion. Each position unit record is spatially linked and matched with the detection area boundary and internal sub-region division to determine the detection sub-region to which each inclusion belongs and the overall detection area. The corresponding region affiliation mark is written into the position unit record. Then, for each marked position unit record, the fields are arranged in the order of inclusion presence identifier, detection area, and three-dimensional spatial coordinates. Multiple position unit records within the same detection area are merged according to spatial position order to form structured inclusion detection result entries. After the results are archived, the inclusion detection results containing inclusion presence / absence determination, distribution area, and precise spatial coordinates are output.

[0118] like Figure 3 The figure illustrates the process of spatially locating inclusions inside a melt based on ultrasonic signals with multiple propagation directions. In the figure, the upper part represents the container wall, and the lower part represents the melt region. Multiple ultrasonic action segments with different propagation directions are formed inside the melt. By setting multiple excitation positions, laser-excited ultrasonic waves enter the melt along different propagation directions, thereby forming multiple action segments in the melt, specifically including action segments of direction 1, direction 2, and direction 3. These action segments extend and intersect within the melt, forming multiple overlapping regions.

[0119] Furthermore, after extracting and matching the scattered signals corresponding to each propagation direction, the candidate signals from different propagation directions are spatially mapped. As shown in the figure, the overlapping area of ​​the action segments corresponding to different propagation directions within the melt is determined as the candidate location area. This candidate location area represents the range of potential inclusions under the joint constraints of multiple propagation paths.

[0120] Based on this, the candidate positioning areas are further narrowed by combining the path constraint information and corresponding signal characteristics of each propagation direction, thus gradually reducing the candidate positioning areas to form the target positioning area. The target positioning area is the region that satisfies the consistency constraint of multiple propagation directions. The consistency of multiple propagation directions includes spatial consistency and feature consistency. Spatial consistency means that the spatial deviation between candidate positions obtained by inversion from different propagation directions is less than a preset spatial threshold. Feature consistency means that the corresponding scattered signals are consistent in phase change trend, spectral shift characteristics, and energy distribution characteristics. Only when both spatial consistency and feature consistency conditions are met are they determined to be the same intrusion target.

[0121] Finally, the center position of the target positioning area is determined as the spatial position of the inclusion in the melt, as shown in the figure.

[0122] Example 2

[0123] Please see Figure 4 Another embodiment of the present invention provides a melt inclusion detection system based on laser ultrasound, comprising: a parameter acquisition module, a detection field construction module, a signal extraction module, and an inclusion detection module;

[0124] The parameter acquisition module is used to set a detection area on the outside of the container wall carrying the melt, acquire the container wall thickness, melt temperature parameters and geometric boundary parameters, and establish a time-varying sound velocity compensation model for the coupling of the container wall and the melt.

[0125] The detection field construction module is used to use pulsed laser to sequentially scan and excite multiple circumferential discrete positions in the detection area, generate ultrasonic waves in the container wall and couple them into the melt, forming a multi-propagation direction detection field covering the interior of the melt.

[0126] The signal extraction module is used to collect time-domain echo signals at discrete locations in each circumferential direction, and to perform wall echo subtraction, flow background suppression and dispersion correction on the time-domain echo signals to extract inclusion scattering signals.

[0127] The impurity detection module is used to perform joint analysis based on the consistency of multiple propagation directions and the phase and spectral characteristics of the impurity scattering signal, to determine the presence and spatial location of impurities in the melt, and to output the impurity detection results.

[0128] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0129] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for detecting inclusions in melts based on laser ultrasound, characterized in that, include: A detection area is set on the outside of the container wall carrying the melt to obtain the container wall thickness, melt temperature parameters and geometric boundary parameters, and to establish a time-varying sound velocity compensation model for the coupling between the container wall and the melt. By using pulsed laser to sequentially scan and excite multiple circumferential discrete positions in the detection area, ultrasonic waves are generated inside the container wall and coupled into the melt, forming a multi-propagation direction detection field covering the interior of the melt. The excitation parameters at each circumferential discrete position are adaptively adjusted according to the time-varying sound velocity compensation model. Time-domain echo signals at discrete locations in each circumferential direction are acquired, and wall echo subtraction, flow background suppression, and dispersion correction are performed on the time-domain echo signals to extract inclusion scattering signals. Based on the consistency of multiple propagation directions and the phase and spectral characteristics of inclusion scattering signals, a joint analysis is performed to determine the presence and spatial location of inclusions in the melt, and the inclusion detection results are output. The method involves sequentially scanning and exciting multiple circumferential discrete positions within the detection area using pulsed lasers to generate ultrasonic waves within the container wall and couple them into the melt, forming a multi-propagation direction detection field covering the interior of the melt. This includes: Using the geometric boundary of the detection area as a constraint, the detection area is divided into multiple excitation position units along the circumferential direction of the outer side of the container wall, and the corresponding circumferential excitation sequence is generated according to the preset orientation order; Each excitation location unit is associated with the corresponding detection sub-region in the time-varying sound speed compensation model, and the propagation path constraint information corresponding to each excitation location unit is extracted to form a one-to-one path constraint set. Based on the path constraint set, the incident position offset order and excitation timing interval of the pulse laser in each excitation position unit are set respectively; According to the circumferential excitation sequence, each excitation position unit is sequentially excited by pulsed laser. After completing one round of circumferential scanning, the incident position offset order is changed to form a new excitation sequence, so that multiple rounds of scanning form an interlaced ultrasonic field with different propagation directions inside the melt. The staggered ultrasonic fields are summarized and rearranged according to their corresponding excitation position units and propagation path constraint information to construct a multi-propagation direction detection field covering the detection area. The step of summarizing and rearranging the staggered ultrasonic fields according to their corresponding excitation position units and propagation path constraint information to construct a multi-propagation direction detection field covering the detection area includes: The ultrasonic propagation trajectories formed by each excitation position unit during each round of interleaved scanning are extracted into independent propagation segments, and then collected according to the excitation round, excitation position unit and corresponding path constraint information to form multiple sets of propagation segments to be rearranged. Based on the orientation order and path constraint information of adjacent excitation position units in each set of propagation segments to be rearranged, the independent propagation segments are sorted by adjacency to form candidate propagation chains classified by propagation direction. According to the incident bias order corresponding to the change of the excitation round, the propagation segments contained in each candidate propagation chain are replaced and rearranged across rounds to obtain multiple directional propagation chains; The directional propagation chains are combined and arranged according to their coverage order within the detection area to establish a correspondence between propagation direction categories and coverage segments of the detection area, and a multi-propagation direction detection field covering the detection area is generated accordingly. The process of acquiring time-domain echo signals at discrete circumferential locations and performing wall echo subtraction, flow background suppression, and dispersion correction on the time-domain echo signals to extract inclusion scattering signals includes: According to the excitation sequence corresponding to each circumferential discrete position, the time-domain echo signals returned after each excitation are continuously acquired, and the time-domain echo signals are grouped and labeled according to the excitation position, excitation round and propagation direction category to form multiple sets of original echo sequences; The wall main echo segments of the original echo sequence are extracted according to the time sequence under the same propagation direction category. The wall main echo segments in the same circumferential discrete position in different excitation rounds are aligned and sorted to form the reference wall echo sequences corresponding to each circumferential discrete position. Using the reference wall echo sequence corresponding to each circumferential discrete position as a reference, the corresponding echo segments in the original echo sequence of the same group are sequentially subtracted by segment replacement and spliced ​​with neighboring residuals to obtain the intermediate echo sequence after removing the main wall echo. For the intermediate echo sequence, echo segments with close time are extracted according to the circumferential correspondence between adjacent excitation cycles, and cross-cycle differential rearrangement and same-direction segment merging are performed on the extracted echo segments to form an echo sequence to be corrected after removing the flow background. Based on the time-varying sound speed compensation model, the echo sequence to be corrected is segmented and propagated according to the corresponding propagation direction category and path constraint information. Time delay anomalous segments, phase anomalous segments and energy mutation segments are extracted from each corrected echo segment. The time delay anomalous segments, phase anomalous segments and energy mutation segments are correlated and aggregated to obtain the inclusion scattering signal. The time-varying sound speed compensation model is used to segment the echo sequence to be corrected according to the corresponding propagation direction category and path constraint information. Time delay anomaly segments, phase anomaly segments, and energy mutation segments are extracted from each corrected echo segment. These segments are then correlated and aggregated to obtain the inclusion scattering signal, including: The echo sequence to be corrected is classified and expanded according to its corresponding propagation direction category. Each echo sequence to be corrected is sequentially mapped to the wall propagation section, the interface transition section and the melt propagation section, forming a correction section sequence corresponding to each propagation direction category. Based on the time-varying sound velocity compensation model, the wall propagation segment, interface transition segment and melt propagation segment in each of the correction segment sequences are aligned and rearranged to form a segmented correction echo chain. The time delay anomaly segment, phase anomaly segment, and energy change segment of each segment of the corrected echo chain are extracted sequentially along the propagation direction. The time delay anomaly segments, phase anomaly segments, and energy mutation segments are correlated and arranged, and the segments that meet the preset correlation order are combined into candidate scattering units. Then, the candidate scattering units in different propagation direction categories are merged accordingly to obtain the inclusion scattering signal.

2. The method for detecting inclusions in melt based on laser ultrasound as described in claim 1, characterized in that, The establishment of the time-varying sound velocity compensation model for the container wall melt coupling includes: The boundary outline of the detection area is set along the outer side of the container wall, and the detection area is divided into grid-like discrete parts to obtain multiple detection sub-regions arranged in directional order; The wall penetration propagation path is extracted along the direction perpendicular to the local surface of the container wall, and the corresponding wall penetration path is segmented and calibrated according to the geometric boundary parameters of each detection sub-region to obtain the local wall thickness parameters and wall thickness continuity parameters of each detection sub-region. The melt temperature parameters are mapped into multiple temperature action zones in a way that corresponds one-to-one with each detection sub-region. Combined with the local wall thickness parameters of each detection sub-region, the propagation time sequence association is marked for each temperature action zone, forming a time sequence correspondence between the detection sub-region and the temperature action zone. Based on the local wall thickness parameters, wall thickness continuity parameters, geometric boundary parameters, and the time sequence correspondence of each detection sub-region, the segmented propagation sequence of laser ultrasound from the outside of the container wall into the inside of the container wall, coupling into the melt through the inside of the container wall, and continuing to propagate in the melt is determined in sequence, forming an interface propagation sequence. Based on the propagation sequences of each interface, the segmented propagation order of different detection sub-regions is combined and organized to establish a time-varying sound velocity compensation model that includes wall thickness segments, temperature-affected segments, and boundary constraints of the detection area.

3. The method for detecting inclusions in melt based on laser ultrasound as described in claim 1, characterized in that, The method utilizes joint analysis of the consistency of multiple propagation directions, phase, and spectral characteristics of inclusion scattering signals to determine the presence and spatial location of inclusions in the melt, and outputs inclusion detection results, including: The inclusion scattering signals are grouped and organized according to the corresponding propagation direction category, excitation position unit and path constraint information to form multiple scattering unit sequences corresponding to different propagation directions, and the occurrence order information in each scattering unit sequence is recorded. Based on the occurrence order information of each scattering unit sequence, the scattering units in different propagation directions are cross-compared. The scattering units that satisfy the temporal connection relationship between adjacent propagation directions are grouped into the same source candidate group, and the scattering units that do not form the same source candidate group are retained as independent candidate units. Phase shift segments, spectral shift segments, and frequency band split segments are extracted from the scattering signals corresponding to the same candidate group and independent candidate unit, respectively. The phase shift segments, spectral shift segments, and frequency band split segments are jointly indexed according to the propagation direction category to form phase spectral feature sequences corresponding to each candidate target. Based on the phase spectrum feature sequence, and combined with the propagation direction category, excitation position unit and path constraint information, the intersection position of each candidate target in the detection area is defined, and the spatial position of the inclusion target in the melt is determined. The determined spatial locations of inclusions are archived, and inclusion detection results containing inclusion presence identifiers, their respective detection areas, and corresponding spatial locations are generated and output.

4. The method for detecting inclusions in melt based on laser ultrasound as described in claim 3, characterized in that, The method of defining the intersection positions of each candidate target within the detection area based on the phase spectrum feature sequence, combined with the propagation direction category, excitation position unit, and path constraint information, to determine the spatial position of the inclusion target in the melt includes: The phase spectrum feature sequences corresponding to each candidate target are expanded according to the propagation direction category to determine the directional action segment corresponding to each candidate target under each propagation direction, forming candidate positioning segment sequences corresponding to different propagation directions. The action segments in each direction are sequentially delivered to the corresponding crossing segments within the detection area, and the connecting boundaries of the action segments in each direction are sequentially adjusted according to the path constraint information to form a set of multiple delivery segments in multiple directions. The delivery segment set is cross-matched pairwise. Delivery segments with overlapping and crossing relationships within the detection area and whose phase spectrum feature sequences maintain the corresponding order are merged into the same candidate intersection range. Each candidate intersection range is then converged step by step according to the number of propagation directions and the order of overlap to form the target positioning area. The distribution order of the target positioning areas and the corresponding candidate target excitation position units is correlated and verified. The positions of the target positioning areas that satisfy the same distribution order constraint are determined, and the determined positions are determined as the spatial positions of the inclusion targets in the melt.

5. The method for detecting inclusions in melt based on laser ultrasound as described in claim 4, characterized in that, The process of archiving the determined spatial locations of inclusions, generating and outputting inclusion detection results that include inclusion presence identifiers, their respective detection areas, and corresponding spatial locations, includes: The spatial positions of each inclusion target in the melt are organized according to the distribution order within the detection area, and position unit records are established for each inclusion target based on the propagation direction category, excitation position unit and path constraint information of the corresponding candidate target. Each location unit record is linked to the detection area boundary and its internal segment division to determine the detection area to which each inclusion target belongs and the position order within the area, and an area affiliation mark is written into the corresponding location unit record. For each location unit record with the write area belonging to the mark, the fields are arranged in the order of the inclusion presence identifier, the detection area to which it belongs and the corresponding spatial location. The location unit records in the same detection area are merged in the order of location to form the corresponding inclusion detection result entries.

6. A melt inclusion detection system based on laser ultrasound, used to implement the melt inclusion detection method based on laser ultrasound as described in any one of claims 1-5, characterized in that, include: The system includes a parameter acquisition module, a detection field construction module, a signal extraction module, and a debris detection module. The parameter acquisition module is used to set a detection area on the outside of the container wall carrying the melt, acquire the container wall thickness, melt temperature parameters and geometric boundary parameters, and establish a time-varying sound velocity compensation model for the coupling of the container wall and the melt. The detection field construction module is used to use pulsed laser to sequentially scan and excite multiple circumferential discrete positions in the detection area, generate ultrasonic waves in the container wall and couple them into the melt, forming a multi-propagation direction detection field covering the interior of the melt. The signal extraction module is used to collect time-domain echo signals at discrete locations in each circumferential direction, and to perform wall echo subtraction, flow background suppression and dispersion correction on the time-domain echo signals to extract inclusion scattering signals. The impurity detection module is used to perform joint analysis based on the consistency of multiple propagation directions and the phase and spectral characteristics of the impurity scattering signal, to determine the presence and spatial location of impurities in the melt, and to output the impurity detection results.

Images