Method, apparatus and equipment for visual reconstruction of multi-bubble phase field in liquid metal

Abstract

This invention relates to the field of nuclear reactor technology, specifically to a method, apparatus, and device for visual reconstruction of the phase field of multiple bubbles within liquid metal. The method includes: acquiring ultrasonic signals received by N ultrasonic transducers during each of T consecutive ultrasonic scans of the liquid metal within a test period; determining a spatiotemporal dynamic response sequence based on the ultrasonic signals received by the N ultrasonic transducers during each ultrasonic scan; characterizing the phase field evolution of the liquid metal within the test period using the spatiotemporal dynamic response sequence; and inputting the spatiotemporal dynamic response sequence into an image reconstruction model to obtain a sequence of phase field distribution maps of the liquid metal within the test period output by the image reconstruction model. This invention captures the dynamic information of the multi-bubble phase field evolution over time using an image reconstruction model based on the spatiotemporal dynamic response sequence, achieving visualized measurement of the dynamic behavior of bubbles within liquid metal.

Description

Technical Field

[0001] This invention relates to the field of nuclear reactor technology, and more specifically, to a method, apparatus, and equipment for visual reconstruction of multi-bubble phase fields within liquid metal. Background Technology

[0002] For next-generation liquid metal-cooled nuclear reactors, the dynamic evolution of bubbles within the liquid metal (including bubble formation, migration, merging, breakup, and instantaneous changes in spatial distribution) affects the reactor system's heat transfer efficiency and safety margin. Two-phase visualization technology for liquid metal is also one of the key technical challenges limiting the development of liquid metal-cooled nuclear reactors. Due to the opacity and corrosiveness of the liquid metal interior, traditional optical methods (such as high-speed camera-based measurement methods) and electrical methods (such as conductivity probe-based measurement methods) cannot achieve visualized measurement of the dynamic behavior of internal bubbles. Furthermore, X-ray methods, due to their complex equipment, the particle shielding properties of liquid metal, and high cost, cannot meet the real-time visualization requirements of complex nuclear reactor systems. Summary of the Invention

[0003] The purpose of this invention is to provide a method, apparatus, and device for visual reconstruction of the phase field of multiple bubbles in liquid metal. Based on the continuous ultrasonic scanning of liquid metal, the ultrasonic transducer receives ultrasonic signals, constructs a spatiotemporal dynamic response sequence, and inputs the spatiotemporal dynamic response sequence into the image reconstruction model to realize the visualization of the dynamic phase field under the flowing bubbles in the liquid metal.

[0004] In a first aspect, the present invention provides a method for visually reconstructing the phase field of multiple bubbles within liquid metal, comprising:

[0005] Based on T consecutive ultrasonic scans of the liquid metal during the test period, the ultrasonic signals received by N ultrasonic transducers during each ultrasonic scan are acquired; where T and N are both integers greater than 1; the N ultrasonic transducers are arranged circumferentially along the outer wall of the liquid metal pipe; during each ultrasonic scan, the N ultrasonic transducers sequentially transmit ultrasonic pulse signals to the liquid metal, and for each transmission event of the ultrasonic transducer, the N ultrasonic transducers receive the ultrasonic signal;

[0006] Based on the ultrasonic signals received by each of the N ultrasonic transducers during each ultrasonic scan, a spatiotemporal dynamic response sequence is determined; the spatiotemporal dynamic response sequence is used to characterize the phase field evolution characteristics of the liquid metal during the test period.

[0007] The spatiotemporal dynamic response sequence is input into the image reconstruction model to obtain the phase field distribution map sequence of the liquid metal during the test period output by the image reconstruction model; the phase field distribution map sequence includes T consecutive phase field distribution maps, and the T phase field distribution maps correspond one-to-one with the scanning time of each of the T ultrasound scans; the image reconstruction model is trained based on the spatiotemporal dynamic response sequence samples and their corresponding phase field distribution map sequence labels.

[0008] In some embodiments, determining the spatiotemporal dynamic response sequence based on the ultrasonic signals received by each of the N ultrasonic transducers during each ultrasonic scan includes:

[0009] Based on the ultrasonic signals received by each of the N ultrasonic transducers during each ultrasonic scan, the phase field characteristic tensor of the liquid metal at the scanning time of each ultrasonic scan is determined.

[0010] The spatiotemporal dynamic response sequence is determined based on the phase-field characteristic tensor of the liquid metal at each ultrasonic scan time.

[0011] As one possible implementation, determining the phase-field characteristic tensor of the liquid metal at each ultrasonic scan moment based on the ultrasonic signals received by the N ultrasonic transducers during each ultrasonic scan includes:

[0012] For any ultrasound scan within the test period, based on the ultrasound signals received by each of the N ultrasound transducers during the ultrasound scan, determine the dynamic response matrix corresponding to the emission event of each ultrasound transducer during the ultrasound scan.

[0013] Based on the dynamic response matrix corresponding to the emission event of each ultrasonic transducer in the ultrasonic scan, the phase field characteristic tensor of the liquid metal at the scanning time of the ultrasonic scan is determined.

[0014] As an example, determining the dynamic response matrix corresponding to the emission event of each of the N ultrasonic transducers during the ultrasonic scan, based on the ultrasonic signals received by each of the N ultrasonic transducers during the ultrasonic scan, includes:

[0015] For each ultrasonic transducer's emission event during the ultrasonic scanning process, based on the ultrasonic signals received by each of the N ultrasonic transducers in the emission event, the amplitude of the ultrasonic signal received by each ultrasonic transducer at each sampling time is determined.

[0016] Based on the amplitude of the ultrasonic signal received by each ultrasonic transducer at each sampling time, the dynamic response matrix corresponding to the transmission event is determined.

[0017] As an example, determining the phase-field characteristic tensor of the liquid metal at the scanning moment of the ultrasonic scan based on the dynamic response matrix corresponding to the emission event of each ultrasonic transducer in the ultrasonic scan includes:

[0018] The N dynamic response matrices in the ultrasonic scan are stacked in the spatial dimension to obtain the phase field characteristic tensor of the liquid metal at the scanning time of the ultrasonic scan.

[0019] As one possible implementation, determining the spatiotemporal dynamic response sequence based on the phase-field characteristic tensor of the liquid metal at each scanning moment of the ultrasonic scan includes:

[0020] The T phase field feature tensors are arranged in chronological order according to the scanning time to obtain the spatiotemporal dynamic response sequence.

[0021] In some embodiments, the image reconstruction model is pre-trained in the following manner:

[0022] Obtain spatiotemporal dynamic response sequence samples and their corresponding phase field distribution map sequence labels;

[0023] The spatiotemporal dynamic response sequence samples are input into the initial image reconstruction model to obtain the phase field distribution map sequence prediction results output by the initial image reconstruction model;

[0024] Based on the phase field distribution map sequence prediction results and the phase field distribution map sequence labels, a first loss value and a second loss value are determined; the first loss value is used to characterize the degree of difference between the phase field distribution map sequence prediction results and the phase field distribution map sequence labels; the second loss value is used to characterize the degree of difference between the amount of change in adjacent frames in the phase field distribution map sequence prediction results and the amount of change in adjacent frames in the phase field distribution map sequence labels.

[0025] The total loss value is determined based on the first loss value and the second loss value;

[0026] Based on the total loss value, the initial image reconstruction model is trained to obtain the trained image reconstruction model.

[0027] A second aspect of the present invention provides a device for visualizing and reconstructing the phase field of multiple bubbles within liquid metal, comprising:

[0028] The acquisition module is used to acquire the ultrasonic signals received by N ultrasonic transducers during each of the T consecutive ultrasonic scans of the liquid metal within the test period; wherein T and N are both integers greater than 1; the N ultrasonic transducers are arranged circumferentially along the outer wall of the liquid metal pipe; during each ultrasonic scan, the N ultrasonic transducers sequentially transmit ultrasonic pulse signals to the liquid metal, and for each transmission event of the ultrasonic transducer, the N ultrasonic transducers receive the ultrasonic signal;

[0029] The determination module is used to determine the spatiotemporal dynamic response sequence based on the ultrasonic signals received by each of the N ultrasonic transducers during each ultrasonic scan; the spatiotemporal dynamic response sequence is used to characterize the phase field evolution characteristics of the liquid metal during the test period;

[0030] The reconstruction module is used to input the spatiotemporal dynamic response sequence into the image reconstruction model to obtain the phase field distribution map sequence of the liquid metal during the test period output by the image reconstruction model; the phase field distribution map sequence includes T consecutive phase field distribution maps, and the T phase field distribution maps correspond one-to-one with the scanning time of each of the T ultrasound scans; the image reconstruction model is trained based on the spatiotemporal dynamic response sequence samples and their corresponding phase field distribution map sequence labels.

[0031] A third aspect of the present invention provides an electronic device, including a processor and a memory storing a computer program, wherein the processor executes the program to implement the visualization and reconstruction method for multi-bubble phase field in liquid metal described in the first aspect above.

[0032] In a fourth aspect, the present invention provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method for visual reconstruction of the multi-bubble phase field within liquid metal as described in the first aspect.

[0033] The method, apparatus, and device for visual reconstruction of multi-bubble phase field within liquid metal provided by this invention have the following beneficial effects:

[0034] This invention constructs a spatiotemporal dynamic response sequence characterizing the phase field evolution of liquid metal during the test period by performing T consecutive ultrasonic scans on liquid metal within the test period and receiving ultrasonic signals from N ultrasonic transducers. Then, based on an image reconstruction model, a mapping is achieved between the spatiotemporal dynamic response sequence and the phase field distribution map sequence. In other words, the dynamic information of the phase field evolution of multiple bubbles over time is captured by the image reconstruction model based on the spatiotemporal dynamic response sequence, realizing the visualization measurement of the dynamic behavior of bubbles in liquid metal and improving the accuracy of phase field image reconstruction in liquid metal. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0036] Figure 1 This is a flowchart illustrating a method for visually reconstructing the phase field of multiple bubbles within liquid metal, as provided in an embodiment of the present invention.

[0037] Figure 2 This is an example diagram showing the layout of the ultrasonic transducer array on the outer periphery of the pipe in an embodiment of the present invention;

[0038] Figure 3 This is an example diagram of the cross-sectional layout of N ultrasonic transducers on the outer periphery of a pipe in an embodiment of the present invention;

[0039] Figure 4 This is a flowchart illustrating another method for visual reconstruction of the multi-bubble phase field within liquid metal provided in an embodiment of the present invention.

[0040] Figure 5 This is a flowchart illustrating another method for visual reconstruction of the multi-bubble phase field within liquid metal provided in an embodiment of the present invention.

[0041] Figure 6 This is a flowchart illustrating another method for visual reconstruction of the multi-bubble phase field within liquid metal provided in an embodiment of the present invention.

[0042] Figure 7 This is a schematic diagram of the structure of a liquid metal multi-bubble phase field visualization and reconstruction device provided in an embodiment of the present invention;

[0043] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0044] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0045] In related technologies, there are limited measurement methods for visualizing multiple bubbles within liquid metal. Existing ultrasonic measurement schemes use artificial intelligence to convert sensor waveforms into one-dimensional vectors and employ fully connected neural networks to map the vectors to images. This significantly increases signal processing complexity and noise intensity, resulting in poor visualization effects, and the number of bubbles has a significant impact on visualization. Furthermore, this method can only obtain single-frame image data, which cannot meet the needs of transient real-time measurement. As for particle imaging methods such as X-ray or neutron beam methods, their high cost, the strong shielding effect of liquid metal against X-ray particles, and the complex and large size of the equipment cannot meet the real-time testing requirements of complex pipeline systems in the nuclear industry.

[0046] When dealing with dynamic flow field measurement, the above-mentioned schemes have the following defects in the existing schemes that combine ultrasound and artificial intelligence: (1) "Static" limitation: The output results are all instantaneous states at a single point in time (such as a static image or an instantaneous average value), lacking the continuity of the time dimension, unable to show the movement trajectory, morphological evolution and velocity information of the bubble group, and unable to meet the analysis needs of the dynamic behavior of the flow field; (2) Destruction of key information structure: The machine learning-based method forcibly flattens the sensor array response data with inherent spatiotemporal structure into a one-dimensional vector, destroying the spatiotemporal correlation between signals, resulting in low model learning efficiency and low reconstruction fidelity.

[0047] To address the aforementioned problems, this invention provides a method, apparatus, and device for visual reconstruction of the phase field of multiple bubbles within liquid metal.

[0048] Figure 1 This is a schematic flowchart illustrating a method for visually reconstructing the phase field of multiple bubbles within liquid metal, provided in an embodiment of the present invention. Figure 1 As shown, the method includes the following steps:

[0049] Step 101: Based on the continuous T ultrasonic scans of the liquid metal during the test period, acquire the ultrasonic signals received by each of the N ultrasonic transducers during each ultrasonic scan.

[0050] Where T and N are both integers greater than 1; N ultrasonic transducers are arranged circumferentially along the outer wall of the liquid metal pipe; during each ultrasonic scan, the N ultrasonic transducers sequentially transmit ultrasonic pulse signals to the liquid metal, and for each ultrasonic transducer transmission event, the N ultrasonic transducers receive the ultrasonic signal.

[0051] In some embodiments, such as Figure 2 As shown, an ultrasonic transducer array containing N ultrasonic transducers is arranged circumferentially on the outer wall of a pipe at a set cross-section of the liquid metal. Figure 3 Here is an example diagram showing the cross-sectional layout of N ultrasonic transducers around the periphery of a pipe, as shown below. Figure 3As shown, N ultrasonic transducers can be evenly arranged around the outer wall of the pipe, and N can be 32.

[0052] In some embodiments, the ultrasonic probe employs a "one-transmitter-multiple-receiver" mode. For each scan, N ultrasonic transducers sequentially transmit ultrasonic pulse signals to the liquid metal, and for each transmission event, all N ultrasonic transducers receive the signal. Specifically, during the current ultrasonic scan, the first ultrasonic transducer is activated and transmits an ultrasonic pulse signal to the liquid metal, while all N ultrasonic transducers simultaneously receive the ultrasonic signal; this continues until all signals have been received. Then, the second ultrasonic transducer is activated and transmits an ultrasonic pulse signal to the liquid metal, while all N ultrasonic transducers simultaneously receive the ultrasonic signal; this continues until all signals have been received. Finally, the third ultrasonic transducer is activated and transmits an ultrasonic pulse signal to the liquid metal, and so on, until all N ultrasonic transducers have transmitted ultrasonic pulse signals and all signals have been received, at which point the next ultrasonic scan is performed.

[0053] In other words, for each transmission event in any given ultrasonic scan, N ultrasonic transducers receive the corresponding ultrasonic signal.

[0054] Step 102: Determine the spatiotemporal dynamic response sequence based on the ultrasonic signals received by each of the N ultrasonic transducers during each ultrasonic scan.

[0055] Among them, the spatiotemporal dynamic response sequence is used to characterize the phase field evolution characteristics of liquid metal during the test period.

[0056] It is understandable that during ultrasonic scanning, due to the different spatial positions of different ultrasonic transducers, the ultrasonic signals received by each transducer based on the emission event are related to the spatial characteristics of the internal phase field of the liquid metal. The ultrasonic signals corresponding to multiple consecutive ultrasonic scans can reflect the changes in the spatial characteristics of the internal phase field of the liquid metal over time. Therefore, the spatiotemporal dynamic response sequence can be determined based on the ultrasonic signals received by N ultrasonic transducers during each ultrasonic scan.

[0057] In some embodiments, a spatiotemporal feature extraction model can be pre-constructed. The ultrasound signals received by each of the N ultrasound transducers during each ultrasound scan are input into the spatiotemporal feature extraction model to obtain the spatiotemporal dynamic response sequence output by the model. The feature extraction network of the spatiotemporal feature extraction model can be built based on a convolutional neural network, an attention mechanism network, other artificial intelligence models, or a combination of multiple artificial intelligence models. The runaway feature extraction model can be jointly trained with an image reconstruction model.

[0058] In other embodiments, for each ultrasonic scan, the phase-field characteristic tensor of the liquid metal at the scanning time of the ultrasonic scan can be determined based on the ultrasonic signals received by each of the N ultrasonic transducers during the ultrasonic scan; the phase-field characteristic tensors corresponding to each of the T consecutive ultrasonic scans are arranged in chronological order to obtain a spatiotemporal dynamic response sequence.

[0059] Step 103: Input the spatiotemporal dynamic response sequence into the image reconstruction model to obtain the phase field distribution sequence of liquid metal in the test period output by the image reconstruction model.

[0060] The phase field distribution map sequence includes T consecutive phase field distribution maps, and each of the T phase field distribution maps corresponds one-to-one with the scanning time of each of the T ultrasound scans. The image reconstruction model is trained based on spatiotemporal dynamic response sequence samples and their corresponding phase field distribution map sequence labels.

[0061] In other words, each frame of the phase field distribution map sequence is the result of a comprehensive judgment based on the phase field characteristics at each ultrasound scan time within the prediction period. By analyzing the contextual information of historical frames, the model can better distinguish between true phase field characteristics and random noise or artifacts in a single frame. For example, a weak signal that persists across multiple frames and conforms to motion patterns is more likely to be identified as a real bubble; while a signal spike that appears only momentarily in a single frame is more likely to be identified as noise and suppressed. This allows for a more reliable and definitive reconstruction of the multi-bubble phase field at each moment.

[0062] In some embodiments, the image reconstruction model can be built based on a convolutional neural network. As an example, the image reconstruction model can be built based on a recurrent convolutional U-Net network, which uses a long short-term memory (LSTM) convolutional unit instead of a standard convolutional unit. U-Net is an excellent static image processing network, adept at processing single two-dimensional images and outputting a corresponding two-dimensional image. However, U-Net itself lacks a temporal concept and memory capability. To meet the needs of real-time measurement, conventional methods first use U-Net to process each frame of video independently, obtaining preliminary results for each frame (e.g., the total area of ​​bubbles in each frame, a one-dimensional value). Then, the sequence of these one-dimensional values ​​is input into an LSTM to analyze its temporal trend—a simple concatenation. This invention fundamentally transforms the core computational unit of U-Net. The convolutional operation originally used to process static spatial information within U-Net is transformed into a recurrent convolutional network capable of processing both spatial and temporal information. This method possesses both the powerful spatial feature extraction and high-fidelity reconstruction capabilities of U-Net and the long-term temporal memory and dynamic pattern learning capabilities of LSTM. This method exhibits a synergistic enhancement effect. By leveraging the memory of LSTM, the model references information from historical frames when reconstructing the current frame. This enables it to effectively distinguish between real physical features and instantaneous random noise, resulting in higher quality and more stable reconstruction results compared to simply using U-Net to process a single frame. The image reconstruction model employs a U-Net encoder-decoder structure, but its internal convolutional modules are all composed of the aforementioned convolutional long short-term memory units. The encoder path is responsible for extracting and compressing spatiotemporal features from the input spatiotemporal dynamic response sequence, while the decoder path is responsible for progressively upsampling the compressed spatiotemporal features and combining them with the high-resolution features passed from the encoder through skip connections, ultimately reconstructing the image.

[0063] The method for visual reconstruction of the phase field of multiple bubbles in liquid metal according to embodiments of the present invention constructs a spatiotemporal dynamic response sequence characterizing the phase field evolution of liquid metal during the test period by performing T consecutive ultrasonic scans on liquid metal during the test period and receiving ultrasonic signals from N ultrasonic transducers. Then, based on an image reconstruction model, a mapping is achieved between the spatiotemporal dynamic response sequence and the phase field distribution map sequence. In other words, the dynamic information of the phase field evolution of multiple bubbles over time is captured by the image reconstruction model based on the spatiotemporal dynamic response sequence, realizing the visual measurement of the dynamic behavior of bubbles in liquid metal and improving the accuracy of phase field image reconstruction in liquid metal.

[0064] Figure 4 This is a schematic flowchart illustrating another method for visualizing and reconstructing the multi-bubble phase field within liquid metal, provided in an embodiment of the present invention. Figure 4 As shown, based on the above embodiments, Figure 1 The implementation process of step 102 includes the following steps:

[0065] Step 401: Based on the ultrasonic signals received by each of the N ultrasonic transducers during each ultrasonic scan, determine the phase field characteristic tensor of the liquid metal at the scanning time of each ultrasonic scan.

[0066] Among them, the phase field characteristic tensor refers to the phase field characteristic information of liquid metal in space.

[0067] In some embodiments, for each ultrasonic scan, an ultrasonic signal matrix S is constructed based on the ultrasonic signals received by each of the N ultrasonic transducers for each emission event during the ultrasonic scan. ij This refers to the ultrasonic signal received by the j-th ultrasonic transducer for the i-th transmission event, and the ultrasonic signal matrix S is used as the phase field characteristic tensor of the liquid metal at the scanning moment of the ultrasonic scan.

[0068] Step 402: Determine the spatiotemporal dynamic response sequence based on the phase field characteristic tensor of the liquid metal at each ultrasonic scan time.

[0069] In some embodiments, the T phase-field feature tensors can be arranged in chronological order of scanning times to obtain a spatiotemporal dynamic response sequence. For example, the scanning time of the first ultrasound scan within the prediction period is t1, the scanning time of the second ultrasound scan is t2, and so on, with the scanning time of the Tth ultrasound scan being t... T The phase field characteristic tensor of the liquid metal at the Tth ultrasonic scan time is S(t T If ), then the spatiotemporal dynamic response sequence is {S(t1), S(t2), ..., S(t... T )}.

[0070] According to an embodiment of the present invention, a method for visually reconstructing the phase field of a multi-bubble liquid metal is proposed. Based on the ultrasonic signals received by N ultrasonic transducers during each ultrasonic scan, the phase field feature tensor of the liquid metal at each scanning moment is determined. Based on the phase field feature tensor of the liquid metal at each scanning moment, a spatiotemporal dynamic response sequence is determined. This invention obtains the phase field feature tensor at each scanning moment, and then determines the spatiotemporal dynamic response sequence based on the obtained T phase field feature tensors, thus obtaining multi-frame dynamic phase field features that can characterize the liquid metal, thereby improving the accuracy of image reconstruction.

[0071] Next, the process of determining the phase field characteristic tensor will be described in detail.

[0072] Figure 5 This is a flowchart illustrating another method for visually reconstructing the phase field of multiple bubbles within liquid metal, provided as an embodiment of the present invention. Figure 5 As shown, based on the above embodiments, Figure 4 The implementation process of step 401 includes the following steps:

[0073] Step 501: For any ultrasound scan within the test period, based on the ultrasound signals received by each of the N ultrasound transducers during the ultrasound scan, determine the dynamic response matrix corresponding to the emission event of each ultrasound transducer during the ultrasound scan.

[0074] The dynamic response matrix corresponding to each ultrasonic transducer's emission event is used to characterize the liquid metal's response to the ultrasonic pulse signal after the ultrasonic transducer emits the signal. Since the ultrasonic signals received by the N ultrasonic transducers are related to the liquid metal's response to the ultrasonic pulse signal for each emission event, the dynamic response matrix corresponding to the emission event can be determined based on the ultrasonic signals received by the N transducers.

[0075] In some embodiments, based on the ultrasonic signals received by each of the N ultrasonic transducers during ultrasonic scanning, the dynamic response matrix corresponding to the emission event of each ultrasonic transducer during ultrasonic scanning is determined, including the following steps:

[0076] Step S1: For each ultrasonic transducer's emission event during the ultrasonic scanning process, based on the ultrasonic signals received by each of the N ultrasonic transducers in the emission event, determine the amplitude of the ultrasonic signal received by each ultrasonic transducer at each sampling time.

[0077] In some embodiments, for each transmission event, multiple consecutive sampling points or multiple discrete sampling points can be determined based on the duration of receiving the ultrasonic signal. Each sampling point corresponds to a sampling time. Based on the above sampling points, the ultrasonic signal received by each ultrasonic transducer is sampled and processed to obtain the ultrasonic signal amplitude at each sampling time. Specifically, for each transmission event, sampling can be performed at a preset frequency within a time period determined by the transmission time and the time when all N ultrasonic transducers have completed signal reception.

[0078] Step S2: Based on the amplitude of the ultrasonic signal received by each ultrasonic transducer at each sampling time, determine the dynamic response matrix corresponding to the transmission event.

[0079] As an example, a dynamic response matrix for a transmission event can be constructed based on the amplitude of the ultrasonic signal received by each ultrasonic transducer at each sampling time, with the ultrasonic transducer as the column index and the sampling time as the row index. Specifically, it is shown in equation (1) below.

[0080] (1);

[0081] in, The scan time of the s-th ultrasound scan; Let be the dynamic response matrix of the i-th emission event in the s-th ultrasound scan; It means The value of the element in the j-th row and k-th column; For the i-th emission event in the s-th ultrasound scan, the j-th ultrasound transducer at the k-th sampling time... The amplitude of the ultrasonic signal; The first sampling time is given, and P represents the total number of sampling points in a single transmission event. This is the Pth sampling time.

[0082] Step 502: Based on the dynamic response matrix corresponding to the emission event of each ultrasonic transducer in ultrasonic scanning, determine the phase field characteristic tensor of the liquid metal at the scanning moment of ultrasonic scanning.

[0083] In some embodiments, for any ultrasonic scan, the dynamic response matrices corresponding to the N emission events during the ultrasonic scan can be stacked in the spatial dimension to obtain the phase field characteristic tensor of the liquid metal at the scanning time of the ultrasonic scan, as shown in the following equation (2).

[0084] (2);

[0085] in, The scanning time of the liquid metal during the s-th ultrasonic scan The phase field characteristic tensor, The dimension is .

[0086] As one possible implementation, the spatiotemporal dynamic response sequence expression is shown in equation (3).

[0087] (3);

[0088] in, It is a spatiotemporal dynamic response sequence. The scanning time of the liquid metal in the first ultrasonic scan The phase field characteristic tensor; The scanning time of the liquid metal in the second ultrasonic scan The phase field characteristic tensor; The time of the Tth ultrasonic scan of the liquid metal The phase field characteristic tensor.

[0089] According to the embodiment of the present invention, the method for visual reconstruction of the phase field of multiple bubbles in liquid metal determines the dynamic response matrix corresponding to the emission event of each ultrasonic transducer during ultrasonic scanning based on the ultrasonic signals received by each of the N ultrasonic transducers during ultrasonic scanning. Based on the dynamic response matrix corresponding to the emission event of each ultrasonic transducer during ultrasonic scanning, the phase field feature tensor of the liquid metal at the scanning time of ultrasonic scanning is determined. This allows the obtained phase field feature tensor to accurately reflect the phase field feature information of the liquid metal in the spatial dimension, thereby enabling the obtained spatiotemporal dynamic response sequence to include phase field feature information in both spatial and temporal dimensions, and realizing the capture of dynamic information of the evolution of the phase field of multiple bubbles over time.

[0090] Next, we will introduce the training process of the image reconstruction model in detail.

[0091] Figure 6 This is a schematic flowchart illustrating another method for visually reconstructing the phase field of multiple bubbles within liquid metal, provided in an embodiment of the present invention. Figure 6 As shown, the image reconstruction model is pre-trained in the following way:

[0092] Step 601: Obtain spatiotemporal dynamic response sequence samples and their corresponding phase field distribution map sequence labels.

[0093] The acquisition method for the spatiotemporal dynamic response sequence samples can be the same as that for the acquisition method for the spatiotemporal dynamic response sequence. The label for the phase field distribution map sequence can be: visualization of the cross-sectional phase field when a bubble rises from liquid metal during the stable stage of X-ray tomography. The cross-section is the same for both X-ray scanning and ultrasonic signals.

[0094] It should be noted that the length of the sequence in the spatiotemporal dynamic response sequence sample is the same as the length of the sequence in the spatiotemporal dynamic response sequence in the above embodiment.

[0095] Step 602: Input the spatiotemporal dynamic response sequence sample into the initial image reconstruction model to obtain the phase field distribution map sequence prediction result output by the initial image reconstruction model.

[0096] Step 603: Determine the first loss value based on the prediction results of the phase field distribution map sequence and the phase field distribution map sequence labels.

[0097] The first loss value characterizes the degree of difference between the predicted phase field distribution map sequence and the phase field distribution map sequence labels. The second loss value characterizes the degree of difference between the amount of change in adjacent frames in the predicted phase field distribution map sequence and the amount of change in adjacent frames in the phase field distribution map sequence labels.

[0098] In some embodiments, the reconstruction loss value at the pixel level for each frame of image can be calculated based on the prediction results of the phase field distribution map sequence and the phase field distribution map sequence label, and then the first loss value can be determined based on the average value of the reconstruction loss value of each frame of image.

[0099] As one possible implementation, the reconstruction loss value at the pixel level for each frame of image can be determined based on the binary entropy loss value and the Dice loss value.

[0100] As an example, the calculation process for the binary entropy loss value is shown in equation (4) below:

[0101] (4);

[0102] in, This refers to the phase field distribution map label of the s-th frame in the phase field distribution map sequence label. This is the prediction result of the phase field distribution map of the s-th frame in the prediction result of the phase field distribution map sequence; is the binary entropy loss value of the prediction result of the phase field distribution map of the s-th frame in the prediction result of the phase field distribution map sequence; P is the total number of pixels in the phase field distribution map; The prediction result of the phase field distribution map of the s-th frame The pixel value of each pixel; The label for the phase field distribution map of the s-th frame. The pixel label value of each pixel.

[0103] As an example, the Dice loss value is used to measure the overlap between the predicted bubble region and the actual bubble region in liquid metal. It is very effective in handling class imbalance problems where the number of bubble pixels is much smaller than the number of background pixels, and can encourage the network to generate bubbles with more complete and accurate shapes. The calculation process of the Dice loss value is shown in the following equation (5):

[0104] (5);

[0105] in, The Dice loss value is the prediction result of the phase field distribution map of the s-th frame in the prediction result of the phase field distribution map sequence. It is a smoothing constant to prevent the denominator from being zero.

[0106] As an example, the process of determining the reconstruction loss value of the phase field distribution map prediction result based on its binary entropy loss value and Dice loss value for each frame of the phase field distribution map prediction result is shown in the following equation (6).

[0107] (6);

[0108] in, is the reconstruction loss value of the predicted phase field distribution map of the s-th frame.

[0109] As an example, the process of determining the first loss value based on the reconstruction loss value of the prediction result of the phase field distribution map of each frame is shown in the following equation (7).

[0110] (7);

[0111] in, This is the first loss value.

[0112] In some embodiments, the process of determining the second loss value based on the prediction results of the phase field distribution map sequence and the phase field distribution map sequence label can be achieved by the following formula (8).

[0113] (8);

[0114] in, is the second loss value; MSE() is the mean squared error function.

[0115] Step 604: Determine the total loss value based on the first loss value and the second loss value.

[0116] In some embodiments, the average of the first loss value and the second loss value can be determined as the total loss value, or the weighted sum of the first loss value and the second loss value can be determined as the total loss value.

[0117] Step 605: Based on the total loss value, train the initial image reconstruction model to obtain the trained image reconstruction model.

[0118] The liquid metal multi-bubble phase field visualization reconstruction method according to embodiments of the present invention considers a first loss value characterizing the pixel-level accuracy of the reconstructed image and a second loss value used to ensure the authenticity and smoothness of changes between consecutive frames when constructing the loss value, so as to improve the efficiency and effect of model training.

[0119] To achieve the above embodiments, the present invention also provides a device for visual reconstruction of the phase field of multiple bubbles in liquid metal.

[0120] Figure 7 This is a schematic diagram of a device for visualizing and reconstructing the phase field of multiple bubbles within liquid metal, provided as an embodiment of the present invention. Figure 7 As shown, the device includes: an acquisition module 710, a determination module 720, and a reconstruction module 730. Wherein:

[0121] The acquisition module 710 is used to acquire the ultrasonic signals received by N ultrasonic transducers during each ultrasonic scan based on T consecutive ultrasonic scans of liquid metal within the test period; where T and N are both integers greater than 1; the N ultrasonic transducers are arranged circumferentially along the outer wall of the liquid metal pipe; during each ultrasonic scan, the N ultrasonic transducers sequentially transmit ultrasonic pulse signals to the liquid metal, and for each transmission event of the ultrasonic transducer, the N ultrasonic transducers receive the ultrasonic signal;

[0122] The determination module 720 is used to determine the spatiotemporal dynamic response sequence based on the ultrasonic signals received by each of the N ultrasonic transducers during each ultrasonic scan; the spatiotemporal dynamic response sequence is used to characterize the phase field evolution characteristics of liquid metal during the test period.

[0123] The reconstruction module 730 is used to input the spatiotemporal dynamic response sequence into the image reconstruction model to obtain the phase field distribution map sequence of liquid metal in the test period output by the image reconstruction model; the phase field distribution map sequence includes a continuous T-frame phase field distribution map, and the T-frame phase field distribution map corresponds one-to-one with the scanning time of each of the T ultrasound scans; the image reconstruction model is trained based on the spatiotemporal dynamic response sequence samples and their corresponding phase field distribution map sequence labels.

[0124] In some embodiments, the determining module 720 is specifically used for:

[0125] Based on the ultrasonic signals received by each of the N ultrasonic transducers during each ultrasonic scan, the phase field characteristic tensor of the liquid metal at the scanning time of each ultrasonic scan is determined.

[0126] Based on the phase field characteristic tensor of liquid metal at each ultrasonic scan time, the spatiotemporal dynamic response sequence is determined.

[0127] As one possible implementation, the determining module 720 is also used for:

[0128] For any ultrasonic scan within the test period, based on the ultrasonic signals received by each of the N ultrasonic transducers during the ultrasonic scan, determine the dynamic response matrix corresponding to the emission event of each ultrasonic transducer during the ultrasonic scan.

[0129] Based on the dynamic response matrix corresponding to the emission event of each ultrasonic transducer in ultrasonic scanning, the phase field characteristic tensor of liquid metal at the scanning moment of ultrasonic scanning is determined.

[0130] As an example, module 720 is also used for:

[0131] For each ultrasonic transducer's emission event during ultrasonic scanning, based on the ultrasonic signals received by each of the N ultrasonic transducers during the emission event, the amplitude of the ultrasonic signal received by each ultrasonic transducer at each sampling time is determined.

[0132] Based on the amplitude of the ultrasonic signal received by each ultrasonic transducer at each sampling time, the dynamic response matrix corresponding to the transmission event is determined.

[0133] As one possible implementation, the determining module 720 is also used for:

[0134] The phase field characteristic tensor of liquid metal at the scanning moment of ultrasonic scanning is obtained by stacking N dynamic response matrices in the spatial dimension.

[0135] In some embodiments, the determining module 720 is further configured to:

[0136] Arrange the T phase field feature tensors in chronological order of scanning time to obtain the spatiotemporal dynamic response sequence.

[0137] In some embodiments, the device further includes a training module 740, which is used for:

[0138] Obtain spatiotemporal dynamic response sequence samples and their corresponding phase field distribution map sequence labels;

[0139] The spatiotemporal dynamic response sequence samples are input into the initial image reconstruction model to obtain the phase field distribution map sequence prediction results output by the initial image reconstruction model;

[0140] Based on the prediction results of the phase field distribution map sequence and the phase field distribution map sequence labels, a first loss value and a second loss value are determined. The first loss value is used to characterize the degree of difference between the prediction results of the phase field distribution map sequence and the phase field distribution map sequence labels. The second loss value is used to characterize the degree of difference between the amount of change in adjacent frames in the prediction results of the phase field distribution map sequence and the amount of change in adjacent frames in the phase field distribution map sequence labels.

[0141] The total loss value is determined based on the first loss value and the second loss value;

[0142] Based on the total loss value, the initial image reconstruction model is trained to obtain the trained image reconstruction model.

[0143] The liquid metal multi-bubble phase field visualization reconstruction device provided by the present invention constructs a spatiotemporal dynamic response sequence characterizing the phase field evolution features of the liquid metal during the test period by performing T consecutive ultrasonic scans on the liquid metal during the test period and receiving ultrasonic signals from N ultrasonic transducers. Then, based on an image reconstruction model, the mapping between the spatiotemporal dynamic response sequence and the phase field distribution map sequence is realized. In other words, the dynamic information of the multi-bubble phase field evolution over time is captured by the image reconstruction model based on the spatiotemporal dynamic response sequence, realizing the visualization measurement of the dynamic behavior of bubbles in the liquid metal and improving the accuracy of phase field image reconstruction in the liquid metal.

[0144] It should be noted that the explanations and descriptions in the above embodiments regarding the visualization and reconstruction method of multi-bubble phase field in liquid metal can also be applied to the visualization and reconstruction device of multi-bubble phase field in liquid metal in the embodiments of the present invention, and will not be repeated here.

[0145] Figure 8 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 8 As shown, the electronic device may include a processor 810, a communication interface 820, a memory 830, and a communication bus 840, wherein the processor 810, the communication interface 820, and the memory 830 communicate with each other via the communication bus 840. The processor 810 can call the computer program in the memory 830 to execute the steps of the visualization and reconstruction method for the multi-bubble phase field in liquid metal provided in the above embodiments.

[0146] For example, the method includes: acquiring the ultrasonic signals received by N ultrasonic transducers during each ultrasonic scan based on T consecutive ultrasonic scans of liquid metal within the test period; where T and N are both integers greater than 1; the N ultrasonic transducers are arranged circumferentially along the outer wall of the liquid metal pipe; during each ultrasonic scan, the N ultrasonic transducers sequentially transmit ultrasonic pulse signals to the liquid metal, and for each transmission event of an ultrasonic transducer, the N ultrasonic transducers receive the ultrasonic signals; based on the ultrasonic signals received by the N ultrasonic transducers during each ultrasonic scan, a spatiotemporal dynamic response sequence is determined; the spatiotemporal dynamic response sequence is used to characterize the phase field evolution characteristics of the liquid metal within the test period; the spatiotemporal dynamic response sequence is input into an image reconstruction model to obtain a phase field distribution map sequence of the liquid metal within the test period output by the image reconstruction model; the phase field distribution map sequence includes T consecutive phase field distribution maps, and the T phase field distribution maps correspond one-to-one with the scanning time of each of the T ultrasonic scans; the image reconstruction model is trained based on the spatiotemporal dynamic response sequence samples and their corresponding phase field distribution map sequence labels.

[0147] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0148] On the other hand, embodiments of the present invention also provide a computer program product, the computer program product including a computer program, the computer program being able to be stored on a computer-readable storage medium, and when the computer program is executed by a processor, the computer is able to perform the steps of the visualization and reconstruction method for multi-bubble phase field in liquid metal provided in the above embodiments.

[0149] On the other hand, embodiments of the present invention also provide a non-transitory computer-readable storage medium having a computer program stored thereon, the computer program being used to cause a processor to execute the visualization and reconstruction method for multi-bubble phase field in liquid metal provided in the above embodiments.

[0150] The non-transitory computer-readable storage medium can be any available medium or data storage device that the processor can access, including but not limited to magnetic memory (e.g., floppy disk, hard disk, magnetic tape, magneto-optical disk (MO)), optical memory (e.g., CD, DVD, BD, HVD), and semiconductor memory (e.g., ROM, EPROM, EEPROM, non-volatile memory (NAND FLASH), solid-state drive (SSD)).

[0151] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0152] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0153] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for visually reconstructing the phase field of multiple bubbles within liquid metal, characterized in that, include: Based on T consecutive ultrasonic scans of the liquid metal during the test period, the ultrasonic signals received by N ultrasonic transducers during each ultrasonic scan are acquired; where T and N are both integers greater than 1; the N ultrasonic transducers are arranged circumferentially along the outer wall of the liquid metal pipe; during each ultrasonic scan, the N ultrasonic transducers sequentially transmit ultrasonic pulse signals to the liquid metal, and for each transmission event of the ultrasonic transducer, the N ultrasonic transducers receive the ultrasonic signal; Based on the ultrasonic signals received by each of the N ultrasonic transducers during each ultrasonic scan, a spatiotemporal dynamic response sequence is determined; the spatiotemporal dynamic response sequence is used to characterize the phase field evolution characteristics of the liquid metal during the test period. The spatiotemporal dynamic response sequence is input into the image reconstruction model to obtain the phase field distribution map sequence of the liquid metal during the test period output by the image reconstruction model; the phase field distribution map sequence includes T consecutive phase field distribution maps, and the T phase field distribution maps correspond one-to-one with the scanning time of each of the T ultrasound scans; the image reconstruction model is trained based on the spatiotemporal dynamic response sequence samples and their corresponding phase field distribution map sequence labels.

2. The method according to claim 1, characterized in that, The determination of the spatiotemporal dynamic response sequence based on the ultrasonic signals received by each of the N ultrasonic transducers during each ultrasonic scan includes: Based on the ultrasonic signals received by each of the N ultrasonic transducers during each ultrasonic scan, the phase field characteristic tensor of the liquid metal at the scanning time of each ultrasonic scan is determined. The spatiotemporal dynamic response sequence is determined based on the phase-field characteristic tensor of the liquid metal at each ultrasonic scan time.

3. The method according to claim 2, characterized in that, The determination of the phase-field characteristic tensor of the liquid metal at each ultrasonic scan moment based on the ultrasonic signals received by the N ultrasonic transducers during each ultrasonic scan includes: For any ultrasound scan within the test period, based on the ultrasound signals received by each of the N ultrasound transducers during the ultrasound scan, determine the dynamic response matrix corresponding to the emission event of each ultrasound transducer during the ultrasound scan. Based on the dynamic response matrix corresponding to the emission event of each ultrasonic transducer in the ultrasonic scan, the phase field characteristic tensor of the liquid metal at the scanning time of the ultrasonic scan is determined.

4. The method according to claim 3, characterized in that, The determination of the dynamic response matrix corresponding to the emission event of each of the N ultrasonic transducers during the ultrasonic scanning process, based on the ultrasonic signals received by each of the N ultrasonic transducers, includes: For each ultrasonic transducer's emission event during the ultrasonic scanning process, based on the ultrasonic signals received by each of the N ultrasonic transducers in the emission event, the amplitude of the ultrasonic signal received by each ultrasonic transducer at each sampling time is determined. Based on the amplitude of the ultrasonic signal received by each ultrasonic transducer at each sampling time, the dynamic response matrix corresponding to the transmission event is determined.

5. The method according to claim 3, characterized in that, The determination of the phase-field characteristic tensor of the liquid metal at the scanning moment of the ultrasonic scan, based on the dynamic response matrix corresponding to the emission event of each ultrasonic transducer in the ultrasonic scan, includes: The N dynamic response matrices in the ultrasonic scan are stacked in the spatial dimension to obtain the phase field characteristic tensor of the liquid metal at the scanning time of the ultrasonic scan.

6. The method according to claim 2, characterized in that, The determination of the spatiotemporal dynamic response sequence based on the phase-field characteristic tensor of the liquid metal at each ultrasonic scan time includes: The T phase field feature tensors are arranged in chronological order according to the scanning time to obtain the spatiotemporal dynamic response sequence.

7. The method according to claim 1, characterized in that, The image reconstruction model is pre-trained in the following manner: Acquire spatiotemporal dynamic response sequence samples and their corresponding phase field distribution map sequence labels; The spatiotemporal dynamic response sequence samples are input into the initial image reconstruction model to obtain the phase field distribution map sequence prediction results output by the initial image reconstruction model; Based on the phase field distribution map sequence prediction results and the phase field distribution map sequence labels, a first loss value and a second loss value are determined; the first loss value is used to characterize the degree of difference between the phase field distribution map sequence prediction results and the phase field distribution map sequence labels; the second loss value is used to characterize the degree of difference between the amount of change in adjacent frames in the phase field distribution map sequence prediction results and the amount of change in adjacent frames in the phase field distribution map sequence labels. The total loss value is determined based on the first loss value and the second loss value; Based on the total loss value, the initial image reconstruction model is trained to obtain the trained image reconstruction model.

8. A device for visualizing and reconstructing the phase field of multiple bubbles within liquid metal, characterized in that, include: The acquisition module is used to acquire the ultrasonic signals received by N ultrasonic transducers during each of the T consecutive ultrasonic scans of the liquid metal within the test period; wherein T and N are both integers greater than 1; the N ultrasonic transducers are arranged circumferentially along the outer wall of the liquid metal pipe; during each ultrasonic scan, the N ultrasonic transducers sequentially transmit ultrasonic pulse signals to the liquid metal, and for each transmission event of the ultrasonic transducer, the N ultrasonic transducers receive the ultrasonic signal; The determination module is used to determine the spatiotemporal dynamic response sequence based on the ultrasonic signals received by each of the N ultrasonic transducers during each ultrasonic scan; the spatiotemporal dynamic response sequence is used to characterize the phase field evolution characteristics of the liquid metal during the test period; The reconstruction module is used to input the spatiotemporal dynamic response sequence into the image reconstruction model to obtain the phase field distribution map sequence of the liquid metal during the test period output by the image reconstruction model; the phase field distribution map sequence includes T consecutive phase field distribution maps, and the T phase field distribution maps correspond one-to-one with the scanning time of each of the T ultrasound scans; the image reconstruction model is trained based on the spatiotemporal dynamic response sequence samples and their corresponding phase field distribution map sequence labels.

9. An electronic device, characterized in that, It includes a processor and a memory storing a computer program, wherein the processor executes the program to implement the method for visual reconstruction of the multi-bubble phase field in liquid metal as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for visual reconstruction of the multi-bubble phase field in liquid metal as described in any one of claims 1 to 7.

Images