Single-point temperature measurement constrained molten pool temperature field ultra-high space-time resolution measurement method

Abstract

The application discloses a kind of single-point temperature measurement constrained molten pool temperature field super-high space-time resolution measurement method, it is related to molten pool temperature field monitoring field.The method includes: based on temperature measurement camera obtains the radiation image of molten pool under two preset wave bands and outputs the low-speed low-resolution two-dimensional temperature field image of molten pool, based on single-point thermometer, obtains the high-speed temperature flow of molten pool;Based on unified system time axis, the low-speed low-resolution two-dimensional temperature field image and high-speed temperature flow are time series calibrated, and the paired data set after time alignment is generated;The low-speed low-resolution two-dimensional temperature field image of adjacent time is input to pre-trained time interpolation frame network, to generate high-frame rate low-resolution temperature field flow with single-point actual measurement temperature as physical constraint;High-frame rate low-resolution temperature field flow is input into super-resolution network, and the high-frame rate high-resolution temperature field image flow is output.The application realizes the purpose of molten pool surface monitoring not dependent on single hardware upgrade, while guaranteeing super-high space-time resolution and temperature physical authenticity.

Description

Technical Field

[0001] This invention relates to the field of molten pool temperature field monitoring technology, and in particular to a method for ultra-high spatiotemporal resolution measurement of molten pool temperature field with single-point temperature measurement constraint. Background Technology

[0002] Metal 3D printing is a core development direction in the current advanced manufacturing field. Its additive manufacturing process involves complex thermophysical processes with extremely high temperature gradients and microsecond-level phase transitions. The molten pool, as the core physical unit bearing energy input and the material melting and solidification, directly determines the density, mechanical properties, and forming accuracy of the final component through its dynamic evolution. Currently, the industry can achieve high-speed observation of features such as keyhole evolution and bubble movement within the molten pool at the level of hundreds of thousands of frames using synchrotron X-ray imaging technology. However, the monitoring capability of the molten pool surface temperature field has long lagged behind, failing to form a spatiotemporally matched collaborative analysis with internal observations, resulting in two common technical bottlenecks:

[0003] First, direct hardware monitoring solutions are constrained by physical limitations and cost: existing hardware upgrade paths either rely on multiple high-speed imaging devices to simultaneously acquire multi-band radiation signals, which is not only expensive but also difficult to completely eliminate spatiotemporal synchronization errors between multiple devices, easily introducing temperature measurement deviations; or they compress multi-band signals into a single camera through a beam splitter, which reduces costs, but is limited by the contradiction between the physical readout speed of the image sensor, exposure time, and signal-to-noise ratio, and the effective frame rate can only be maintained at a maximum of 10. 4 At the fps level, the observation frequency is orders of magnitude lower than that of synchrotron radiation, resulting in a microsecond-level evolutionary blind spot that cannot be covered.

[0004] Second, pure algorithmic reconstruction schemes lack guarantees of physical authenticity: existing computer vision-based frame interpolation and super-resolution algorithms generally suffer from the problem of "emphasizing image features while neglecting physical laws." Traditional frame interpolation algorithms only generate intermediate frames based on pixel motion compensation, without introducing thermodynamic constraints on temperature evolution, which easily leads to artifacts that violate the logic of heat conduction and have no practical measurement value. Although new sensing schemes such as event cameras can capture microsecond-level brightness changes, they can only obtain the relative rate of change and cannot be directly mapped to the absolute measured temperature. Most deep learning super-resolution schemes only optimize the spatial resolution of a single frame and do not perform deep reconstruction of the continuous thermal evolution process between frames, and the temporal resolution is still locked by the hardware sampling frequency. A few algorithms that introduce physical constraints rely heavily on high-frequency two-dimensional measured temperatures that strictly correspond to spatial positions. However, in the microscopic transient melt pool scenario, obtaining such true values ​​is an industry bottleneck that has not yet been overcome, and the schemes cannot be implemented.

[0005] In summary, the industry urgently needs a solution for monitoring the surface of molten pools that does not rely on a single hardware upgrade while ensuring ultra-high spatiotemporal resolution and accurate temperature physics. Summary of the Invention

[0006] This invention provides a method for ultra-high spatiotemporal resolution measurement of molten pool temperature field under single-point temperature measurement constraint, which aims to achieve molten pool surface monitoring without relying on a single hardware upgrade, while ensuring ultra-high spatiotemporal resolution and temperature physical accuracy. The technical solution is as follows:

[0007] In a first aspect, embodiments of the present invention provide a method for ultra-high spatiotemporal resolution measurement of molten pool temperature field under single-point temperature measurement constraint, characterized in that it includes:

[0008] The system acquires radiation images of the molten pool in two preset bands using a temperature measuring camera and outputs a low-speed, low-resolution two-dimensional temperature field image of the molten pool, as well as a high-speed temperature flow of the molten pool using a single-point thermometer. The temperature measuring camera and the single-point thermometer are respectively arranged on the imaging optical path and the temperature measuring optical path of the molten pool radiation light separated by a beam splitter.

[0009] Based on a unified system time axis, the low-speed, low-resolution two-dimensional temperature field image and the high-speed temperature flow are time-aligned to generate a time-aligned paired dataset; the paired dataset includes low-speed, low-resolution two-dimensional temperature field images at adjacent times and single-point measured temperatures within the corresponding time interval.

[0010] The low-speed, low-resolution two-dimensional temperature field images at adjacent time points are input into a pre-trained temporal interpolation network to generate a high-frame-rate, low-resolution temperature field flow using the measured temperature at a single point as a physical constraint.

[0011] A high-frame-rate, low-resolution temperature field stream is input into a pre-trained super-resolution network, which outputs a high-frame-rate, high-resolution temperature field image stream.

[0012] In some embodiments of the present invention, the step of performing time-series calibration on the low-speed, low-resolution two-dimensional temperature field image and the high-speed temperature flow to generate a time-aligned paired dataset includes:

[0013] In the initial state, the first reference mean and the first standard deviation of the low-speed, low-resolution two-dimensional temperature field image signal sequence are calculated, and the second reference mean and the second standard deviation of the high-speed temperature flow signal sequence are calculated.

[0014] A first noise benchmark is established for a low-speed, low-resolution two-dimensional temperature field image based on the first benchmark mean and the first standard deviation, and a second noise benchmark is established for a high-speed temperature flow based on the second benchmark mean and the second standard deviation.

[0015] Recursive calculations are performed on the low-speed, low-resolution two-dimensional temperature field image and the high-speed temperature flow data to determine a first statistic for the low-speed, low-resolution two-dimensional temperature field image and a second statistic for the high-speed temperature flow data. The first statistic represents the degree to which the low-speed, low-resolution two-dimensional temperature field image deviates from a first noise reference, and the second statistic represents the degree to which the high-speed temperature flow data deviates from a second noise reference.

[0016] When the first statistic is greater than the statistical threshold, the current first statistical moment is determined, and based on the current first moment, the moment when the first statistic is 0 is determined to be the starting moment of the low-speed, low-resolution two-dimensional temperature field image; when the second statistic is greater than the statistical threshold, the current second statistical moment is determined, and based on the current second moment, the moment when the second statistic is 0 is determined to be the starting moment of the temperature flow data.

[0017] The starting time of the low-speed, low-resolution two-dimensional temperature field image and the starting time of the high-speed temperature flow data are mapped to the zero point of the unified system time axis.

[0018] In some embodiments of the present invention, the step of inputting the low-speed, low-resolution two-dimensional temperature field images of adjacent time moments into a pre-trained temporal interpolation network, and generating a high-frame-rate, low-resolution temperature field flow using the single-point measured temperature as a physical constraint, includes:

[0019] Low-speed, low-resolution two-dimensional temperature field images at adjacent time points are input into a pre-trained temporal interpolation network to generate at least one predicted temperature field image; the predicted temperature field image is located between the low-speed, low-resolution two-dimensional temperature field images at adjacent time points; wherein, the physical constraint mechanism of the temporal interpolation network is to use the measured temperature at a single point as a supervision signal and optimize the network parameters by minimizing the physical constraint loss function.

[0020] The predicted temperature field image is input into a pre-trained temporal interpolation network, which outputs a high frame rate, low-resolution temperature field flow.

[0021] In some embodiments of the present invention, the training process of the time interpolation network includes:

[0022] Low-speed, low-resolution two-dimensional temperature field images at adjacent time points are input into a pre-trained temporal interpolation network to generate intermediate frame prediction images, and the average temperature and peak temperature of the entire image are extracted from the intermediate frame prediction images.

[0023] The average temperature and peak temperature of the entire map are converted into predicted effective temperatures that are consistent with and comparable to the measured temperature at a single point.

[0024] The parameters of the time interpolation network are updated based on the deviation between the predicted effective temperature and the measured temperature at a single point at the same time.

[0025] In some embodiments of the present invention, the temporal interpolation network loss function is expressed by the following formula:

[0026]

[0027]

[0028] in, This represents the physical constraint loss value. Represents the deviation measurement operator. This indicates the measured temperature at a single point. Indicates the predicted effective temperature. This represents a function operator that reflects the mapping relationship between the predicted effective temperature and the average and peak temperatures. This indicates the average temperature of the molten pool. This indicates the peak temperature of the molten pool.

[0029] In some embodiments of the present invention, the step of inputting a high frame rate, low resolution temperature field stream into a pre-trained super-resolution network and outputting a high frame rate, high resolution temperature field image stream includes:

[0030] The input high frame rate low resolution temperature field flow is convolved using a convolution operator to obtain shallow features.

[0031] The shallow features are input into multiple consecutive mixed attention groups to obtain deep features;

[0032] The shallow features and the deep features are residually joined to obtain fused features;

[0033] The fused features are expanded in spatial dimension using a subpixel convolution operator to generate a high frame rate, high resolution temperature field image stream.

[0034] In some embodiments of the present invention, the temperature measuring camera and the single-point thermometer are used to measure the target central region of the molten pool.

[0035] Secondly, the present invention also provides a high spatiotemporal resolution measurement device for the temperature field of a molten pool with single-point temperature measurement constraint, comprising:

[0036] The acquisition module is used to acquire radiation images of the molten pool in two preset bands based on a temperature measuring camera and output a low-speed, low-resolution two-dimensional temperature field image of the molten pool, as well as to acquire a high-speed temperature flow of the molten pool based on a single-point thermometer; wherein the temperature measuring camera and the single-point thermometer are respectively arranged on the imaging optical path and the temperature measuring optical path of the molten pool radiation light separated by a beam splitter.

[0037] The pairing module is used to perform time-series calibration on the low-speed, low-resolution two-dimensional temperature field image and the high-speed temperature flow based on a unified system time axis, and generate a time-aligned paired dataset; the paired dataset includes low-speed, low-resolution two-dimensional temperature field images at adjacent time points and single-point measured temperatures within the corresponding time interval.

[0038] The prediction module is used to input the low-speed, low-resolution two-dimensional temperature field images of adjacent time moments into a pre-trained temporal interpolation network, and generate a high-frame-rate, low-resolution temperature field flow with the measured temperature at a single point as a physical constraint.

[0039] The output module is used to input a high frame rate, low resolution temperature field stream into a pre-trained super-resolution network and output a high frame rate, high resolution temperature field image stream.

[0040] Thirdly, the present invention also provides an electronic device, comprising: a processor and a memory;

[0041] The memory stores a computer-readable program that can be executed by the processor;

[0042] When the processor executes the computer-readable program, it implements the steps in the ultra-high spatiotemporal resolution measurement method for the molten pool temperature field with single-point temperature measurement constraint as described above.

[0043] Fourthly, the present invention also provides a computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the steps in the single-point temperature-constrained ultra-high spatiotemporal resolution measurement method for the molten pool temperature field as described above.

[0044] The beneficial effects of the technical solutions provided by some embodiments of the present invention include at least the following: First, acquiring radiation images of the molten pool in two preset bands based on a temperature measuring camera and outputting a low-speed, low-resolution two-dimensional temperature field image of the molten pool; and acquiring a high-speed temperature flow of the molten pool based on a single-point thermometer; thereby replacing the expensive high-resolution ultra-high-speed area array thermal imager. Subsequently, based on a unified system time axis, the low-speed, low-resolution two-dimensional temperature field image and the high-speed temperature flow are time-aligned to generate a time-aligned paired dataset; the paired dataset includes low-speed, low-resolution two-dimensional temperature field images at adjacent times and images within the corresponding time interval. Single-point temperature measurement eliminates time axis misalignment caused by hardware response differences. Subsequently, the low-speed, low-resolution two-dimensional temperature field images of adjacent time points are input into a pre-trained temporal interpolation network to generate a high-frame-rate, low-resolution temperature field flow with the single-point temperature measurement as a physical constraint. This breaks through the physical sampling limit of the camera without increasing hardware costs. Finally, the high-frame-rate, low-resolution temperature field flow is input into a pre-trained super-resolution network to output a high-frame-rate, high-resolution temperature field image flow, achieving a synergistic improvement in spatiotemporal resolution. This enables ultra-high-resolution monitoring of the molten pool surface temperature field that matches the synchrotron radiation observation frequency. Attached Figure Description

[0045] 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 some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 A system architecture diagram of an embodiment of the ultra-high spatiotemporal resolution measurement method for molten pool temperature field under single-point temperature measurement constraint provided by the present invention;

[0047] Figure 2 A flowchart of an embodiment of the ultra-high spatiotemporal resolution measurement method for molten pool temperature field under single-point temperature measurement constraint provided by the present invention;

[0048] Figure 3 for Figure 2 In step S201, a schematic diagram of the arrangement of the temperature measuring camera and the single-point thermometer is shown.

[0049] Figure 4 for Figure 2 A flowchart of a method according to an embodiment of step S202;

[0050] Figure 5 for Figure 2 A flowchart of a method according to an embodiment of step S203;

[0051] Figure 6This is a schematic diagram of an embodiment of the ultra-high spatiotemporal resolution measurement device for molten pool temperature field with single-point temperature measurement constraint provided by the present invention.

[0052] Figure 7 This is a schematic diagram of the operating environment of an embodiment of the electronic device provided by the present invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0054] It should be noted that the ultra-high spatiotemporal resolution measurement method for the molten pool temperature field under single-point temperature measurement constraint provided in this application is generally executed by the terminal equipment. Correspondingly, the ultra-high spatiotemporal resolution measurement device for the molten pool temperature field under single-point temperature measurement constraint is generally installed in the terminal equipment.

[0055] Figure 1 An exemplary system architecture is shown that can be applied to the single-point temperature measurement constraint of the molten pool temperature field ultra-high spatiotemporal resolution measurement method or the single-point temperature measurement constraint of the molten pool temperature field ultra-high spatiotemporal resolution measurement device of this application.

[0056] like Figure 1 As shown, the system architecture may include: terminal device 11 and server 12. Terminal device 11 and server 12 can communicate via a network, which serves as the medium for providing communication links between the various units. The network may include various types of wired or wireless communication links, such as: wired communication links including fiber optic cables, twisted-pair cables, or coaxial cables; and wireless communication links including Bluetooth communication links, Wi-Fi communication links, or microwave communication links.

[0057] It should be noted that the terminal device 11 and the server 12 can be either hardware or software. When the terminal device 11 and the server 12 are hardware, they can be implemented as a distributed server cluster consisting of multiple servers, or as a single server. When the terminal device 11 and the server 12 are software, they can be implemented as multiple software programs or software modules (for example, to provide distributed services), or as a single software program or software module; no specific limitations are made here.

[0058] The terminal device of this application can be equipped with various communication client applications, such as video recording applications, video playback applications, voice interaction applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0059] A terminal device can be either hardware or software. When the terminal device is hardware, it can be various terminal devices with a display screen, including but not limited to smartphones, tablets, laptops, and desktop computers. When the terminal device is software, it can be installed on the terminal devices listed above. It can be implemented as multiple software programs or software modules (e.g., used to provide distributed services) or as a single software program or software module; no specific limitation is made here.

[0060] When the terminal device is hardware, it can also be equipped with a display device and a camera. The display device can be any device capable of displaying information, and the camera is used to capture video streams. For example, the display device can be a cathode ray tube display (CR), a light-emitting diode display (LED), an e-ink screen, a liquid crystal display (LCD), a plasma display panel (PDP), etc. Users can use the display device on the terminal device to view displayed text, images, videos, and other information.

[0061] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is for illustrative purposes only. Depending on implementation needs, there can be any number of terminal devices, networks, and servers.

[0062] The following will be combined with the appendix Figure 2 This application provides a detailed description of the ultra-high spatiotemporal resolution measurement method for the molten pool temperature field under single-point temperature measurement constraint provided in the embodiments of this application. Specifically, the ultra-high spatiotemporal resolution measurement device for the molten pool temperature field under single-point temperature measurement constraint in the embodiments of this application can be... Figure 1 The terminal device shown.

[0063] Please see Figure 2 This document presents a flowchart illustrating a method for ultra-high spatiotemporal resolution measurement of the molten pool temperature field under single-point temperature measurement constraint, as described in an embodiment of this application. Figure 2 As shown, the method described in this application embodiment may include the following steps:

[0064] S201. Acquire radiation images of the molten pool in two preset bands based on a temperature measuring camera and output a low-speed, low-resolution two-dimensional temperature field image of the molten pool, and acquire a high-speed temperature flow of the molten pool based on a single-point thermometer; wherein, the temperature measuring camera and the single-point thermometer are respectively arranged on the imaging optical path and the temperature measuring optical path of the molten pool radiation light separated by a beam splitter.

[0065] It should be noted that the temperature measuring camera and single-point thermometer are part of the beam splitting and synchronous acquisition system. This system is positioned along the propagation path of the molten pool's thermal radiation and is used to split, guide, and collect the radiated light from the molten pool. Figure 3 As shown, the system includes a dichroic beam splitter, a synchronization controller, a temperature measuring camera, and a single-point thermometer. 101 represents the molten pool, 102 represents the beam splitter, 103 represents the temperature measuring camera, 104 represents the single-point thermometer, and 105 represents the synchronization controller. The dichroic beam splitter is positioned on the radiation path near the molten pool, spatially separating the radiated light from the molten pool into an imaging path and a temperature measuring path. The synchronization controller connects both the temperature measuring camera and the single-point thermometer. The temperature measuring camera is positioned at the reflected light path of the imaging path to receive the imaging beam separated by the dichroic beam splitter. The single-point thermometer is positioned at the transmitted light path of the temperature measuring path to receive the temperature measuring beam separated by the dichroic beam splitter.

[0066] Furthermore, the dichroic beam splitter, a semi-reflective mirror or a dichroic mirror, is placed at 45° to the measuring optical axis. It divides the radiation information of the molten pool into two parts: one part is transmitted to the temperature measuring camera through the reflection imaging optical path, and the other part is transmitted to the single-point thermometer through the transmission temperature measuring optical path. The temperature measuring camera is a temperature field measuring camera based on the principle of dual-band colorimetric thermometry. It can be a dual-spectrum or multi-spectrum camera or a temperature measuring camera combined with a beam splitter optical path. It acquires radiation images of the molten pool in two specific bands, and outputs a low-speed, low-resolution two-dimensional temperature field image of the molten pool after calculation and calibration. The single-point thermometer is a colorimetric thermometer with a sampling frequency not less than 10 times the frequency of the temperature measuring camera.

[0067] S202. Based on the unified system time axis, perform time-series calibration on the low-speed, low-resolution two-dimensional temperature field image and the high-speed temperature flow to generate a time-aligned paired dataset; the paired dataset includes low-speed, low-resolution two-dimensional temperature field images at adjacent times and single-point measured temperatures within the corresponding time interval.

[0068] S203. Input the low-speed, low-resolution two-dimensional temperature field images of adjacent time moments into a pre-trained temporal interpolation network, and generate a high-frame-rate, low-resolution temperature field flow with the measured temperature at a single point as a physical constraint.

[0069] S204. Input the high frame rate, low resolution temperature field stream into the pre-trained super-resolution network and output a high frame rate, high resolution temperature field image stream.

[0070] In this embodiment, firstly, a low-speed, low-resolution two-dimensional temperature field image of the molten pool is acquired using a temperature-measuring camera under two preset wavelengths, and a high-speed temperature flow of the molten pool is output. Secondly, a high-speed temperature flow of the molten pool is acquired using a single-point thermometer, thus replacing the expensive high-resolution ultra-high-speed area array thermal imager. Then, based on a unified system time axis, the low-speed, low-resolution two-dimensional temperature field image and the high-speed temperature flow are time-aligned to generate a time-aligned paired dataset. This paired dataset includes low-speed, low-resolution two-dimensional temperature field images at adjacent time points and the single-point measured temperature within the corresponding time interval, eliminating time axis misalignment caused by hardware response differences. Next, the low-speed, low-resolution two-dimensional temperature field images at adjacent time points are input into a pre-trained time interpolation network, using the single-point measured temperature as a physical constraint to generate a high-frame-rate, low-resolution temperature field flow, breaking through the physical sampling limit of the camera without increasing hardware costs. Finally, the high-frame-rate, low-resolution temperature field flow is input into a pre-trained super-resolution network, outputting a high-frame-rate, high-resolution temperature field image flow, achieving a synergistic improvement in spatiotemporal resolution, thereby completing ultra-high-resolution monitoring of the molten pool surface temperature field that matches the synchrotron radiation observation frequency.

[0071] Due to inherent delay differences in hardware startup response, data transmission, and exposure processing between the temperature measuring camera and the single-point thermometer, relying solely on the hardware trigger signal of the synchronization controller cannot achieve strict physical time alignment between the two data streams. To address this issue, in some embodiments of the present invention, the low-speed, low-resolution two-dimensional temperature field image and the temperature flow are time-aligned to generate a time-aligned paired dataset. Please refer to [link to relevant documentation]. Figure 4 ,include:

[0072] S401. In the initial state, calculate the first reference mean and the first standard deviation of the low-speed, low-resolution two-dimensional temperature field image signal sequence, and calculate the second reference mean and the second standard deviation of the high-speed temperature flow signal sequence.

[0073] S402. Establish a first noise benchmark for a low-speed, low-resolution two-dimensional temperature field image based on the first benchmark mean and the first standard deviation, and establish a second noise benchmark for a high-speed temperature flow based on the second benchmark mean and the second standard deviation.

[0074] S403. Perform recursive calculations on the real-time low-speed, low-resolution two-dimensional temperature field image and the high-speed temperature flow data respectively to determine the first statistic of the low-speed, low-resolution two-dimensional temperature field image and the second statistic of the high-speed temperature flow data; wherein, the first statistic is the degree to which the low-speed, low-resolution two-dimensional temperature field image deviates from the first noise reference, and the second statistic is the degree to which the high-speed temperature flow data deviates from the second noise reference.

[0075] S404. When the first statistic is greater than the statistical threshold, determine the current first statistical moment, and backtrack based on the current first moment to determine the moment when the first statistic is 0 as the start moment of the low-speed, low-resolution two-dimensional temperature field image; when the second statistic is greater than the statistical threshold, determine the current second statistical moment, and backtrack based on the current second moment to determine the moment when the second statistic is 0 as the start moment of the temperature flow data.

[0076] S405. Map the start time of the low-speed, low-resolution two-dimensional temperature field image and the start time of the high-speed temperature flow data to a unified system time axis zero point.

[0077] In this embodiment, an adaptive time alignment method based on CUSUM statistics is adopted. To address the time axis misalignment problem caused by the different start-up delay and response characteristics of the temperature measuring camera and the single-point thermometer, the synchronization controller alone cannot be used for time calibration. Therefore, an adaptive alignment algorithm based on bilateral cumulative sum statistics is adopted.

[0078] First, within the silent time window after the system starts acquiring data but before the laser is turned on, the data processing system performs statistical analysis on the signals acquired by the temperature camera and the single-point thermometer, respectively, and calculates the first reference mean of the low-speed, low-resolution two-dimensional temperature field image signal sequence. and the first standard deviation And calculate the second reference mean of the temperature flow signal sequence. Second standard deviation A first noise benchmark for the low-speed, low-resolution two-dimensional temperature field image is established based on the first benchmark mean and the first standard deviation, and a second noise benchmark for the temperature flow is established based on the second benchmark mean and the second standard deviation. Then, the system samples each data point input in real time. Perform recursive operations to calculate the positive cumulative sum statistic. This statistic reflects the cumulative degree of signal deviation from the background reference, and its calculation formula is as follows:

[0079]

[0080] in, express Cumulative deviation value at time step, initial state ; express The signal sample value at time; For the drift parameter, in this embodiment, we take... ,in The sensitivity coefficient is 0.5, with a preferred value of 0.5.

[0081] System-defined decision thresholds , here set to Real-time monitoring during the calculation process. ,when When, determine The moment the laser is activated is recorded, and the alarm time at this moment is recorded. However, due to the lag in the accumulation process, the system performs a backtracking operation, from... Search forward and backward continuously to find Mark the last time a value is 0 as the start time. The calculations show that, through the above steps, the actual start time of laser activation is detected in the signal streams of the temperature camera and the single-point thermometer, and this start time is mapped to the zero point of the unified system time axis, thereby eliminating hardware startup delay and response differences in the multimodal data acquisition process and achieving precise alignment of the two signals in the time dimension.

[0082] It should be noted that the methods for timing calibration are not limited to those described above.

[0083] Existing deep learning-based video frame interpolation techniques are mainly used to improve video smoothness. The generated intermediate frames only guarantee pixel-level visual continuity and lack physical meaning. In molten pool temperature field monitoring, relying solely on image features for frame interpolation leads to generated temperature values ​​deviating from the actual thermophysical process, failing to meet the needs of scientific measurement. Furthermore, due to hardware cost limitations, it is difficult to directly obtain high-frequency two-dimensional measured temperatures for supervised training. Therefore, in some embodiments of this invention, the low-speed, low-resolution two-dimensional temperature field images at adjacent time points are input into a pre-trained temporal frame interpolation network to generate a high-frame-rate, low-resolution temperature field flow using the single-point measured temperature as a physical constraint. Please refer to [link to relevant documentation]. Figure 5 ,include:

[0084] S501. Input the low-speed, low-resolution two-dimensional temperature field images at adjacent time points into a pre-trained temporal interpolation network to generate at least one frame of predicted temperature field image; the predicted temperature field image is located between the low-speed, low-resolution two-dimensional temperature field images at adjacent time points; wherein, the physical constraint mechanism of the temporal interpolation network is to use the measured temperature at a single point as a supervision signal and optimize the network parameters by minimizing the physical constraint loss function.

[0085] S502. Input the predicted temperature field image into a pre-trained temporal interpolation network to output a high frame rate, low resolution temperature field flow.

[0086] In this embodiment, low-speed, low-resolution two-dimensional temperature field image data of the molten pool, generated through numerical simulation based on a thermal conduction physical model, is input into a time interpolation network for initial training. This allows the model to extract and learn the basic optical flow and motion morphology features of the molten pool evolution. The pre-trained model parameters are then loaded, and the model is further optimized using real temperature field image data captured by a temperature measuring camera. During this process, a loss function with physical constraints is used, which calculates the difference between the predicted image temperature value and the actual temperature value measured by a single-point thermometer at the same time. The optimized model parameters are then loaded, and the molten pool temperature field images from two adjacent frames output by the temperature measuring camera are received to generate a high frame rate, low-resolution temperature field image stream between the two frames.

[0087] Specifically, during the training phase, two adjacent frames of the low-speed, low-resolution two-dimensional temperature field image are obtained. The data is input into an interpolation network to generate predicted images at intermediate time points, and segmented high-speed temperature data at corresponding time points are extracted. Temperature feature values ​​are used as physical constraints. The predicted image is optimized by minimizing the physical constraint loss function. The physical constraint loss function is constructed as follows: extract the pixel-level grayscale matrix of the intermediate frame of the prediction and map it to temperature values, and calculate the average temperature of the entire image. With peak temperature Since a single-point thermometer cannot directly measure the average or peak temperature of the entire image, but rather measures an effective temperature that reflects local thermal radiation characteristics, this invention constructs an effective temperature mapping operator to establish comparability between the predicted image and the true value at a single point. Subsequently, the effective temperature mapping operator is used to extract the corresponding predicted effective temperature from the generated prediction frame. The deviation is measured by a deviation metric function compared to the true value collected by a single-point thermometer. This deviation metric function characterizes the difference between the predicted effective value and the true value. In practical applications, absolute deviation, quadratic deviation, or a weighted combination thereof can be used depending on the requirements. The physical constraint loss formula is as follows:

[0088]

[0089]

[0090] in, This represents the physical constraint loss value. Represents the deviation measurement operator. This indicates the measured temperature at a single point. Indicates the predicted effective temperature. This represents a function operator that reflects the mapping relationship between the predicted effective temperature and the average and peak temperatures. This indicates the average temperature of the molten pool. This indicates the peak temperature of the molten pool.

[0091] Due to hardware limitations, the spatial resolution of the original acquired images and the interpolated images is limited. The high frame rate, low resolution image stream is input into a pre-trained spatial super-resolution network (such as a HAT network) to reconstruct a high frame rate, high resolution final temperature field image stream. Specifically, the process of inputting the high frame rate, low resolution temperature field stream into the pre-trained super-resolution network and outputting a high frame rate, high resolution temperature field image stream includes:

[0092] The input high frame rate low resolution temperature field flow is convolved using a convolution operator to obtain shallow features.

[0093] The shallow features are input into multiple consecutive mixed attention groups to obtain deep features;

[0094] The shallow features and the deep features are residually joined to obtain fused features;

[0095] The fused features are expanded in spatial dimension using a subpixel convolution operator to generate a high frame rate, high resolution temperature field image stream.

[0096] In this embodiment, the spatial super-resolution module is implemented using a depth residual architecture. The input is a low-resolution temperature field image generated by the physically constrained frame interpolation module. The specific reconstruction steps are as follows:

[0097] The first step is to use the convolution operator. The input image is mapped to a feature space for shallow feature extraction, using convolution operators. Map the input image to the feature space:

[0098]

[0099] in, To extract shallow features, enter A continuous Hybrid Attention Group (RHAG) is used, with its core computation being the Hybrid Attention Block (HAB), which combines window self-attention and channel attention to extract deep features from the image.

[0100] For the The calculation logic for each feature block is as follows:

[0101]

[0102] Within HAB, hybrid attention It is achieved through the following formula:

[0103]

[0104] in, These represent the query, key, and value matrices generated from the input features, respectively. This is a relative position offset used to enhance the spatial correlation of the melt pool morphology. It incorporates depth features. shallow features After long-range residual fusion, the sampled layer is entered. Subpixel convolution operators are then used. Achieving spatial dimension expansion:

[0105]

[0106] For r-fold super-resolution, the sub-pixel convolution operator will continuously convert r values ​​in the input feature tensor. 2 The channels are rearranged in a preset order into a spatial grid that is r times the size of the input image, so that the height and width of the output image are each r times the size of the input image.

[0107] This allows for the conversion of high-frame-rate, low-resolution temperature field images into high-frame-rate, high-resolution image streams, which can then be compared and studied with internal features of the same frame rate to analyze the behavioral characteristics of the molten pool.

[0108] The present invention has the following beneficial effects:

[0109] (1) A deep learning-based temporal frame interpolation and spatial super-resolution concatenation algorithm was adopted, combined with an adaptive temporal alignment mechanism, which broke through the hardware physical limit of the traditional thermal radiation imaging system that "high frame rate and high resolution cannot be achieved at the same time", and for the first time achieved a temperature measurement frame rate of not less than 10. 5 Ultra-high spatiotemporal resolution temperature field monitoring with a frequency of Hz and a spatial resolution of not less than 1024×1024.

[0110] (2) The optical radiation information of the molten pool is obtained by adopting the principle of multi-band colorimetric temperature measurement, which effectively avoids the measurement error caused by the dynamic change of emissivity due to the drastic phase change on the surface of the molten pool during the metal additive manufacturing process. The emissivity of the object being measured does not need to be pre-calibrated, and the temperature distribution information with strong anti-interference ability and high robustness can be obtained.

[0111] (3) The method of using the high-frequency effective measured temperature collected by the single-point thermometer as the physical constraint condition and constructing a multi-modal physical constraint joint loss function to optimize the neural network was adopted, which significantly improved the quantitative numerical temperature measurement accuracy of the temperature field image and ensured that the generated temperature field flow has rigorous scientific measurement value.

[0112] (4) A lightweight multimodal hardware architecture combining a dual-color beam splitter optical path with a conventional high-speed temperature measurement camera and a single-point thermometer is adopted. Through software and hardware collaboration and algorithm compensation, it replaces the extremely expensive ultra-high-speed, ultra-high-resolution area array infrared thermal imager, greatly reducing the overall equipment deployment cost of the system and making the device structure simpler.

[0113] (5) Multi-source data fusion and spatiotemporal joint super-resolution technology were adopted to generate a high-frequency high-definition temperature field image stream that is aligned with the actual microsecond-level evolution process. This enables panoramic online monitoring of the transient changes in the temperature field of the molten pool during laser processing, providing extremely reliable data support for in-depth analysis of the coupling mechanism of surface thermal behavior and internal defect evolution in the additive manufacturing process, as well as the thermodynamic and fluid dynamic performance characterization of the molten pool.

[0114] The following are embodiments of the apparatus described in this application, which can be used to execute the embodiments of the method described in this application. For details not disclosed in the apparatus embodiments of this application, please refer to the embodiments of the method described in this application.

[0115] Please see Figure 6 This illustration shows a schematic diagram of a high spatiotemporal resolution measurement device for a molten pool temperature field with single-point temperature measurement constraint provided in an exemplary embodiment of this application, hereinafter referred to as device 6. Device 6 can be implemented as all or part of a terminal device through software, hardware, or a combination of both. Device 6 includes:

[0116] The acquisition module 610 is used to acquire radiation images of the molten pool in two preset bands based on a temperature measuring camera and output a low-speed, low-resolution two-dimensional temperature field image of the molten pool, and to acquire a high-speed temperature flow of the molten pool based on a single-point thermometer; wherein the temperature measuring camera and the single-point thermometer are respectively arranged on the imaging optical path and the temperature measuring optical path of the molten pool radiation light separated by a beam splitter.

[0117] The pairing module 620 is used to perform time-series calibration on the low-speed, low-resolution two-dimensional temperature field image and the high-speed temperature flow based on a unified system time axis, and generate a time-aligned pairing dataset; the pairing dataset includes low-speed, low-resolution two-dimensional temperature field images at adjacent time points and single-point measured temperatures within the corresponding time interval.

[0118] Prediction module 630 is used to input the low-speed, low-resolution two-dimensional temperature field images of adjacent time moments into a pre-trained temporal interpolation network to generate a high-frame-rate, low-resolution temperature field flow with the single-point measured temperature as a physical constraint.

[0119] The output module 640 is used to input a high frame rate, low resolution temperature field stream into a pre-trained super-resolution network and output a high frame rate, high resolution temperature field image stream.

[0120] This application also provides a computer storage medium that can store multiple instructions, which are adapted to be loaded and executed by a processor as described above. Figure 2 The method steps of the illustrated embodiment can be found in the following documentation for detailed execution. Figure 2 The specific details of the illustrated embodiments will not be elaborated here.

[0121] This application also provides a computer program product that stores at least one instruction, which is loaded and executed by the processor to implement the ultra-high spatiotemporal resolution measurement method for molten pool temperature field under single-point temperature measurement constraint as described in the above embodiments.

[0122] Please see Figure 7 This is a schematic diagram of the structure of a terminal device provided in an embodiment of this application. Figure 7 As shown, the terminal device 700 may include: at least one processor 701, at least one network interface 704, user interface 703, memory 705, and at least one communication bus 702.

[0123] The communication bus 702 is used to enable communication between these components.

[0124] The user interface 703 may include a display screen and a camera. Optionally, the user interface 703 may also include a standard wired interface and a wireless interface.

[0125] The network interface 704 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0126] The processor 701 may include one or more processing cores. The processor 701 connects to various parts within the terminal device 700 using various interfaces and lines, and performs various functions and processes data of the terminal device 700 by running or executing instructions, programs, code sets, or instruction sets stored in the memory 705, and by calling data stored in the memory 705. Optionally, the processor 701 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 701 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display on the screen; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 701 and may be implemented as a separate chip.

[0127] The memory 705 may include random access memory (RAM) or read-only memory. Optionally, the memory 705 may include a non-transitory computer-readable storage medium. The memory 705 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 705 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 705 may also be at least one storage device located remotely from the aforementioned processor 701. Figure 7 As shown, the memory 705, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and application programs.

[0128] exist Figure 7In the terminal device 700 shown, the user interface 703 is mainly used to provide an input interface for the user and to obtain the user's input data; while the processor 701 can be used to call the application program stored in the memory 705 and specifically execute, such as Figure 2 The method shown can be referred to for details. Figure 2 As shown, it will not be elaborated further here.

[0129] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory, or random access memory, etc.

[0130] The above description discloses only preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. Therefore, equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A single-point temperature measurement constrained molten pool temperature field ultra-high spatiotemporal resolution measurement method, characterized in that, include: The system acquires radiation images of the molten pool in two preset bands using a temperature measuring camera and outputs a low-speed, low-resolution two-dimensional temperature field image of the molten pool, as well as a high-speed temperature flow of the molten pool using a single-point thermometer. The temperature measuring camera and the single-point thermometer are respectively arranged on the imaging optical path and the temperature measuring optical path of the molten pool radiation light separated by a beam splitter. Based on a unified system time axis, the low-speed, low-resolution two-dimensional temperature field image and the high-speed temperature flow are time-aligned to generate a time-aligned paired dataset; the paired dataset includes low-speed, low-resolution two-dimensional temperature field images at adjacent times and single-point measured temperatures within the corresponding time interval. The low-speed, low-resolution two-dimensional temperature field images at adjacent time points are input into a pre-trained temporal interpolation network to generate a high-frame-rate, low-resolution temperature field flow using the measured temperature at a single point as a physical constraint. A high-frame-rate, low-resolution temperature field stream is input into a pre-trained super-resolution network, which outputs a high-frame-rate, high-resolution temperature field image stream.

2. The single point temperature measurement constrained molten pool temperature field ultra-high spatio-temporal resolution measurement method according to claim 1, characterized in that, The step of performing time-series calibration on the low-speed, low-resolution two-dimensional temperature field image and the high-speed temperature flow to generate a time-aligned paired dataset includes: In the initial state, the first reference mean and the first standard deviation of the low-speed, low-resolution two-dimensional temperature field image signal sequence are calculated, and the second reference mean and the second standard deviation of the high-speed temperature flow signal sequence are calculated. A first noise benchmark is established for a low-speed, low-resolution two-dimensional temperature field image based on the first benchmark mean and the first standard deviation, and a second noise benchmark is established for a high-speed temperature flow based on the second benchmark mean and the second standard deviation. Recursive calculations are performed on the low-speed, low-resolution two-dimensional temperature field image and the high-speed temperature flow data to determine a first statistic for the low-speed, low-resolution two-dimensional temperature field image and a second statistic for the high-speed temperature flow data. The first statistic represents the degree to which the low-speed, low-resolution two-dimensional temperature field image deviates from a first noise reference, and the second statistic represents the degree to which the high-speed temperature flow data deviates from a second noise reference. When the first statistic is greater than the statistical threshold, the current first statistical moment is determined, and based on the current first moment, the moment when the first statistic is 0 is determined to be the starting moment of the low-speed, low-resolution two-dimensional temperature field image; when the second statistic is greater than the statistical threshold, the current second statistical moment is determined, and based on the current second moment, the moment when the second statistic is 0 is determined to be the starting moment of the temperature flow data. The starting time of the low-speed, low-resolution two-dimensional temperature field image and the starting time of the high-speed temperature flow data are mapped to the zero point of the unified system time axis.

3. The single point temperature measurement constrained molten pool temperature field ultra-high spatio-temporal resolution measurement method according to claim 1, characterized in that, The step of inputting the low-speed, low-resolution two-dimensional temperature field images of adjacent time moments into a pre-trained temporal interpolation network, and generating a high-frame-rate, low-resolution temperature field flow using the single-point measured temperature as a physical constraint, includes: Low-speed, low-resolution two-dimensional temperature field images at adjacent time points are input into a pre-trained temporal interpolation network to generate at least one predicted temperature field image; the predicted temperature field image is located between the low-speed, low-resolution two-dimensional temperature field images at adjacent time points; wherein, the physical constraint mechanism of the temporal interpolation network is to use the measured temperature at a single point as a supervision signal and optimize the network parameters by minimizing the physical constraint loss function. The predicted temperature field image is input into a pre-trained temporal interpolation network, which outputs a high frame rate, low-resolution temperature field flow.

4. The method for ultra-high spatiotemporal resolution measurement of molten pool temperature field under single-point temperature measurement constraint according to claim 3, characterized in that, The training process of the time interpolation network includes: Low-speed, low-resolution two-dimensional temperature field images at adjacent time points are input into a pre-trained temporal interpolation network to generate intermediate frame prediction images, and the average temperature and peak temperature of the entire image are extracted from the intermediate frame prediction images. The average temperature and peak temperature of the entire map are converted into predicted effective temperatures that are consistent with and comparable to the measured temperature at a single point. The parameters of the time interpolation network are updated based on the deviation between the predicted effective temperature and the measured temperature at a single point at the same time.

5. The method for ultra-high spatiotemporal resolution measurement of molten pool temperature field under single-point temperature measurement constraint according to claim 4, characterized in that, The loss function of the time interpolation network is expressed by the following formula: in, This represents the physical constraint loss value. Represents the deviation measurement operator. This indicates the measured temperature at a single point. Indicates the predicted effective temperature. This represents a function operator that reflects the mapping relationship between the predicted effective temperature and the average and peak temperatures. This indicates the average temperature of the molten pool. This indicates the peak temperature of the molten pool.

6. The method for ultra-high spatiotemporal resolution measurement of molten pool temperature field under single-point temperature measurement constraint according to claim 1, characterized in that, The process of inputting a high-frame-rate, low-resolution temperature field stream into a pre-trained super-resolution network and outputting a high-frame-rate, high-resolution temperature field image stream includes: The input high frame rate low resolution temperature field flow is convolved using a convolution operator to obtain shallow features. The shallow features are input into multiple consecutive mixed attention groups to obtain deep features; The shallow features and the deep features are residually joined to obtain fused features; The fused features are expanded in spatial dimension using a subpixel convolution operator to generate a high frame rate, high resolution temperature field image stream.

7. The method for ultra-high spatiotemporal resolution measurement of molten pool temperature field under single-point temperature measurement constraint according to claim 1, characterized in that, The temperature measuring camera and the single-point thermometer are used to measure the same target central area of ​​the molten pool.

8. A high spatiotemporal resolution measurement device for molten pool temperature field with single-point temperature measurement constraint, characterized in that, include: The acquisition module is used to acquire radiation images of the molten pool in two preset bands based on a temperature measuring camera and output a low-speed, low-resolution two-dimensional temperature field image of the molten pool, as well as to acquire a high-speed temperature flow of the molten pool based on a single-point thermometer; wherein the temperature measuring camera and the single-point thermometer are respectively arranged on the imaging optical path and the temperature measuring optical path of the molten pool radiation light separated by a beam splitter. The pairing module is used to perform time-series calibration on the low-speed, low-resolution two-dimensional temperature field image and the high-speed temperature flow based on a unified system time axis, and generate a time-aligned paired dataset; the paired dataset includes low-speed, low-resolution two-dimensional temperature field images at adjacent time points and single-point measured temperatures within the corresponding time interval. The prediction module is used to input the low-speed, low-resolution two-dimensional temperature field images of adjacent time moments into a pre-trained temporal interpolation network, and generate a high-frame-rate, low-resolution temperature field flow with the measured temperature at a single point as a physical constraint. The output module is used to input a high frame rate, low resolution temperature field stream into a pre-trained super-resolution network and output a high frame rate, high resolution temperature field image stream.

9. An electronic device, characterized in that, include: Processor and memory; The memory stores a computer-readable program that can be executed by the processor; When the processor executes the computer-readable program, it implements the steps in the ultra-high spatiotemporal resolution measurement method for molten pool temperature field with single-point temperature measurement constraint as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more programs, which can be executed by one or more processors to implement the steps in the method for ultra-high spatiotemporal resolution measurement of molten pool temperature field under single-point temperature measurement constraint as described in any one of claims 1-7.

Images