Method and system for optimizing the forming process of a high flux tube with a twisted sintered porous layer

Abstract

This invention discloses a method and system for optimizing the forming process of a twisted sintered porous layer high-throughput tube, relating to the field of intelligent manufacturing technology. The method includes: acquiring a pre-defined three-dimensional pore structure model of the high-throughput tube; quantifying the target pore features into pore feature tensors; and obtaining an acoustic field control program through inverse mapping using a pre-trained process parameter generation network. During the metal powder laying process, the powder laying device and ultrasonic field parameters are simultaneously controlled to form a pore precursor structure in the powder prefabrication layer. The powder prefabrication layer with the pore precursor structure and the spiral twisted base tube are then sintered. After sintering, actual pore structure data is collected, structural deviations are calculated, and the process parameter generation network is iteratively updated. This invention solves the technical problem that traditional manufacturing processes cannot accurately achieve three-dimensional control of the porous layer pore structure, resulting in significant deviations between the formed pore structure and the target requirements. It achieves the technical effects of precise forming of the porous layer pore structure and dynamic control of the entire process.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing technology, and in particular to a method and system for optimizing the forming process of twisted sintered porous layer high-flux tubes. Background Technology

[0002] The molding process of twisted sintered porous layer high-throughput tubes is crucial for improving the performance of heat exchange equipment and ensuring its stable operation under high heat flux density conditions. Optimizing the process strategy is the core key to achieving precise molding of porous structures and improving product quality. Existing technologies mostly employ the traditional powder-layout-sintering process to prepare these high-throughput tubes, relying on fixed process parameters to complete powder laying and sintering, which has played a certain role in the preparation of simple layered gradient porous structures. However, with the increasing demand for precise three-dimensional programming of porous structures, traditional processes have revealed many limitations in the preparation of this type of high-throughput tube. Traditional processes lack dynamic process strategy optimization capabilities and cannot adapt to the helical twisted base tube morphology to achieve three-dimensional control of the pore structure, resulting in large deviations between the molded pores and the target, poor controllability, and difficulty in meeting the requirements for high-performance preparation and precise quality control of high-throughput tubes. Summary of the Invention

[0003] This application provides a method and system for optimizing the forming process of twisted sintered porous layer high-flux tubes, which solves the technical problem that traditional preparation processes cannot accurately achieve three-dimensional control of the porous layer pore structure and that the formed pore structure deviates significantly from the target requirements.

[0004] The first aspect of this application provides a method for optimizing the forming process of a twisted sintered porous layer high-throughput tube. The method includes: obtaining a preset three-dimensional pore structure model of the target product, wherein the preset three-dimensional pore structure model defines the target pore features of the metal powder porous layer of the high-throughput tube at different spatial locations using a digital model, and quantizing the target pore features into a computable pore feature tensor; inputting the pore feature tensor into a pre-trained process parameter generation network, inversely mapping and outputting an acoustic field control program that can realize the target pore features, wherein the acoustic field control program includes a spatiotemporal sequence instruction for controlling a programmable focused ultrasonic field; During the metal powder laying process, the motion parameters of the powder laying device and the output parameters of the programmable focused ultrasonic field are synchronously coordinated and controlled according to the spatiotemporal sequence instructions in the acoustic field control program. This allows the programmable focused ultrasonic field to perform real-time irradiation modulation on the powder preform layer according to a preset program, thereby forming a porous precursor structure within the powder preform layer. The powder preform layer with the porous precursor structure and the spiral twisted base tube are then sintered. After sintering, real-time pore structure data of the actual molded product is collected, and structural deviation is calculated with the pore feature tensor. Based on the calculated structural deviation, the process parameter generation network is iteratively updated.

[0005] A second aspect of this application provides a forming process optimization system for a twisted sintered porous layer high-throughput tube. The system includes: a pore feature tensor acquisition module for acquiring a preset three-dimensional pore structure model of the target product. The preset three-dimensional pore structure model defines the target pore features of the metal powder porous layer of the high-throughput tube at different spatial locations using a digital model, and quantifies the target pore features into a computable pore feature tensor; and an acoustic field control program output module for inputting the pore feature tensor into a pre-trained process parameter generation network, inversely mapping it, and outputting an acoustic field control program that can realize the target pore features. The acoustic field control program includes a spatiotemporal sequence command for controlling a programmable focused ultrasonic field. The precursor structure construction module is used to synchronously coordinate and control the motion parameters of the powder laying device and the output parameters of the programmable focused ultrasonic field according to the spatiotemporal sequence instructions in the acoustic field control program during the metal powder laying process. This allows the programmable focused ultrasonic field to perform real-time irradiation modulation on the powder preform layer according to a preset program, thereby forming a porous precursor structure within the powder preform layer. The process parameter generation network update module is used to sinter the powder preform layer with the porous precursor structure and the spiral twisted base tube. After sintering, it collects real-time pore structure data of the actual formed product, calculates the structural deviation with the pore feature tensor, and iteratively updates the process parameter generation network based on the calculated structural deviation.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application obtains a target three-dimensional pore structure model of a twisted high-throughput tube and quantifies the pore features into a computable tensor. The tensor is input into a pre-trained network to generate an acoustic field control program. During powder spreading, the powder spreading device and the programmable focused ultrasonic field are synchronously coordinated to form a pore precursor. After sintering, actual pore data is collected to calculate the deviation and iteratively update the network. Combined with virtual verification and acoustic feedback, adaptive process control is achieved, thereby accurately realizing the programming and construction of the three-dimensional pore structure of the metal powder porous layer of the high-throughput tube. This makes the forming process of the twisted sintered porous layer high-throughput tube more accurate and controllable. The pore structure of the formed product is highly consistent with the target requirements. The process can also be continuously optimized through iteration, which greatly improves the forming quality and performance of the high-throughput tube. It achieves the technical effect of precise forming of porous layer pore structure and dynamic control of the entire process. Attached Figure Description

[0007] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0008] Figure 1 This is a flowchart illustrating the optimization method for the forming process of the twisted sintered porous layer high-flux tube provided in the embodiments of this application.

[0009] Figure 2 This is a schematic diagram of the molding process optimization system for the twisted sintered porous layer high-throughput tube provided in the embodiments of this application.

[0010] Figure labeling: 1. Pore feature tensor acquisition module; 2. Acoustic field control program output module; 3. Pore precursor structure construction module; 4. Process parameter generation network update module. Detailed Implementation

[0011] This application provides a method and system for optimizing the forming process of twisted sintered porous layer high-flux tubes, which solves the technical problem that traditional preparation processes cannot accurately achieve three-dimensional control of the porous layer pore structure and that the formed pore structure deviates significantly from the target requirements.

[0012] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0013] It should be noted that the terms "first," "second," etc., in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to such processes, methods, products, or devices.

[0014] Example 1, as Figure 1 As shown, an optimized forming process method for a twisted sintered porous layer high-flux tube is provided, wherein the method includes: A preset three-dimensional pore structure model of the target product is obtained. The preset three-dimensional pore structure model defines the target pore features of the metal powder porous layer of the high-throughput tube at different spatial locations using a digital model, and quantifies the target pore features into a computable pore feature tensor.

[0015] In this embodiment, the twisted sintered porous high-flux tube is a metal tube with a spiral twisted structure as the base tube and a gradient porous metal powder layer sintered on the outer surface. It is a core heat exchange component in phase change heat exchange equipment such as evaporators and reboilers to achieve efficient heat transfer and adapt to harsh working conditions with high heat flux density.

[0016] Specifically, based on the actual operating conditions of high-flux tubes with twisted sintered porous layers in high-heat-flux heat exchange equipment such as evaporators and reboilers, and the preset performance indicators such as heat transfer coefficient and thermal shock resistance, the target pore characteristics that the metal powder porous layer needs to meet at different locations in three-dimensional space are determined. These include the porosity, average pore diameter, and pore connectivity of the large-diameter bottom region close to the spiral twisted base tube, and the porosity, average pore diameter, and orientation angle of the pore channels along the spiral direction of the base tube in the small-diameter surface region far from the base tube. Then, using three-dimensional porous structure digital modeling methods, and utilizing the parametric modeling methods of professional software such as ANSYS Fluent and COMSOL Multiphysics, combined with the determined target pore characteristics at each spatial location, a preset three-dimensional pore structure model that can accurately characterize the global pore feature distribution of the metal powder porous layer is established. This model can completely reflect the pore structure state of the porous layer at different spatial coordinates.

[0017] Next, the established three-dimensional porous structure model is spatially discretized. Based on the actual molding accuracy requirements of the high-throughput tube's porous metal powder layer, the voxel unit division scale is selected, typically using 50μm-200μm cubic voxels. The three-dimensional space of the porous metal powder layer is divided into regularly arranged voxel units of this size, with each voxel unit corresponding to a unique three-dimensional spatial coordinate, completing the global discretization of the porous layer's three-dimensional space. For each discretized voxel unit, key target pore feature parameters characterizing the pore structure are extracted from the established three-dimensional porous structure model. The extracted parameter types are selected based on the high-throughput tube's performance requirements such as heat transfer and mechanical bonding, and may include core parameters such as porosity, average pore diameter, pore connectivity, and pore channel orientation angle, achieving accurate extraction of target pore features for each voxel unit.

[0018] Then, all target pore feature parameters of each voxel unit are normalized and quantized. The normalization method is used to eliminate the dimensional differences between different types of pore feature parameters. The linear normalization method can be selected to map the parameter values ​​to the standard value range of [0,1]. The calculation method is quantized value = (original parameter value - minimum parameter value) / (maximum parameter value - minimum parameter value). The quantized value of each target pore feature parameter in each voxel unit is obtained, so that different types of pore feature parameters are formed into a digital form that can be calculated uniformly.

[0019] Furthermore, the pore feature tensor is a multidimensional tensor constructed by quantizing the pore feature parameters of each voxel unit in the three-dimensional space of the porous layer according to the characterization rules of the distribution of physical quantities in multidimensional space. It can transform discrete quantized pore feature values ​​into tensor data forms that can be recognized and calculated by computers, thus fully characterizing the distribution law of pore features in three-dimensional space. The selection of pore feature parameters corresponds to the action mechanism of the programmable focused ultrasonic field on powder particles, so that the pore features of each voxel unit can be mapped to a controllable sound field energy distribution and action strategy, thereby establishing the correlation between pore structure and sound field control.

[0020] Taking the extraction of three types of target pore feature parameters—porosity, average pore size, and pore connectivity—as an example, the three-dimensional space of the porous layer of metal powder is discretized into M×N×K voxel units, where M, N, and K are the number of voxel units along the x, y, and z axes of the three-dimensional space, respectively. The constructed pore feature tensor is T∈R^(M×N×K×3), where R represents the real number field. Each element T(i,j,k,p) in the tensor corresponds to the quantized value of the p-th target pore feature parameter of the i-th voxel unit along the x-axis, j-th voxel unit along the y-axis, and k-th voxel unit along the z-axis. When p is 1, it is the quantized value of porosity; when p is 2, it is the quantized value of average pore size; and when p is 3, it is the quantized value of pore connectivity.

[0021] Finally, following the arrangement order of the three-dimensional voxel units, the pore feature parameters of all voxel units after normalization and quantization are ordered and combined according to the tensor dimension construction rules to form the pore feature tensor corresponding to the preset three-dimensional pore structure model. This pore feature tensor not only characterizes the spatial pore structure distribution but also serves as the target constraint input for the subsequent process parameter generation network, driving the inverse generation of the sound field control program and realizing the mapping from the target pore structure to the sound field control strategy. Furthermore, the pore feature tensor maintains the spatial adjacency relationships and pore connectivity constraints between voxel units in the three-dimensional space, thereby supporting the coordinated control of the subsequent sound field within a continuous spatial region and improving the overall consistency and controllability of the pore structure construction.

[0022] The pore feature tensor is input into a pre-trained process parameter generation network, which is then inversely mapped and outputs an acoustic field control program that can achieve the target pore features. The acoustic field control program includes spatiotemporal sequence instructions for controlling the programmable focused ultrasonic field.

[0023] Optionally, the pore feature tensor is first input into a pre-trained process parameter generation network to perform inverse mapping operation, generating a set of meta-instructions that specifies the staged control strategy of the acoustic field focal array in different regions of the three-dimensional powder bed and can induce the preset self-organization behavior of powder particles. Then, this set is compiled into a spatiotemporal sequence instruction that can be recognized and executed by the programmable focused ultrasonic field underlying device, forming an acoustic field control program containing such instructions. This step will be explained in detail later.

[0024] During the metal powder laying process, the motion parameters of the powder laying device and the output parameters of the programmable focused ultrasonic field are synchronously coordinated and controlled according to the spatiotemporal sequence instructions in the acoustic field control program, so that the programmable focused ultrasonic field performs real-time irradiation modulation on the powder preform layer according to the preset program, so as to form a porous precursor structure in the powder preform layer.

[0025] In this embodiment, the programmable focused ultrasonic field is an array of multiple ultrasonic transducers whose phase and output amplitude can be independently adjusted. By programming and controlling the emission timing and parameters of each transducer, a focused ultrasonic field with dynamic generation, movement and precise control of energy distribution is generated in a three-dimensional powder bed for non-contact guidance of the directional movement and self-assembly of metal powder particles.

[0026] In one embodiment of this application, during the metal powder laying process, the acoustic field control program is analyzed and a synchronous control signal is generated to drive the powder laying device to lay a powder preform layer on the surface of the spiral twisted base tube. At the same time, the programmable focused ultrasonic field is dynamically tracked and applied to the designated area to form the corresponding pore precursor structure in situ. This step will also be described in detail later.

[0027] The powder preform layer with a porous precursor structure and the spiral twisted base tube are sintered. After sintering, real-time pore structure data of the actual molded product are collected, and structural deviation is calculated with the pore feature tensor. The process parameter generation network is iteratively updated based on the calculated structural deviation.

[0028] In this embodiment, the spiral twisted base tube is a metal base tube with a spiral twisted shape on its surface or as a whole. It serves as a support substrate for preparing a porous layer on its surface, which can enhance fluid disturbance and improve heat exchange or mass transfer efficiency. It is commonly used in high-throughput fluid processing devices.

[0029] Specifically, the powder preform with a porous precursor structure and the spiral twisted base tube are first sintered together and placed in a vacuum or protective atmosphere sintering furnace. The sintering process is carried out according to the preset heating curve, holding temperature and holding time, so that metallurgical bonding is formed between the metal powder particles and between the powder and the base tube surface. After cooling in the furnace, a twisted sintered porous high-throughput tube with a preset three-dimensional porous structure is obtained.

[0030] Next, after sintering, non-destructive testing is performed on the actual molded product to obtain and quantify the actual pore structure data. The actual structural deviation between the actual pore structure and the target pore feature tensor is calculated. This deviation, the target pore feature tensor, and the actual sound field control program are used to form an incremental training sample. This sample is then used to fine-tune and iteratively update the process parameter generation network. This step will also be explained in detail later.

[0031] Furthermore, the method provided in this application embodiment includes: A training dataset containing multiple sets of sample pairs is constructed, wherein each set of sample pairs contains a training pore feature tensor as the network input and a pre-calibrated training sound field control program as the expected output of the network; the training pore feature tensor is used as the input and the training sound field control program is used as the expected output, and the neural network architecture is used for iterative training until a high-dimensional nonlinear mapping relationship is established between the complex three-dimensional pore configuration and the dynamic sound field programming strategy required to realize the configuration, thereby generating a process parameter generation network.

[0032] Optionally, a training dataset containing multiple sample pairs is first constructed. Each sample pair consists of a training pore feature tensor as network input and a pre-calibrated training acoustic field control program as the expected output of the network. The method for constructing the training pore feature tensor is consistent with the method for constructing the pore feature tensor of the target product. Based on a preset three-dimensional pore structure model, the three-dimensional space is discretized into regular voxel units. The pore feature parameters of each voxel unit are extracted, normalized, and quantized, and then combined into a multidimensional tensor in spatial order. The training acoustic field control program is obtained through multiple molding experiments and calibration. For different three-dimensional pore configurations, the focal array scanning path, energy temporal control waveform, and mode switching logic of the programmable focused ultrasound field are repeatedly debugged and compiled into spatiotemporal sequence instructions executable by the underlying device, serving as the expected output of the sample pair. Next, the training dataset is preprocessed. The training pore feature tensor is uniformly normalized to the [0,1] interval, and the training acoustic field control program is parameterized and encoded into a fixed-length sequence, which is divided into a training set and a validation set in an 8:2 ratio.

[0033] Secondly, an adapted neural network model is constructed. The training objective of the process parameter generation network is to achieve the inverse mapping from the target pore structure to the sound field control strategy. This mapping relationship has strong nonlinearity and spatial coupling characteristics, unlike conventional parameter prediction. It is necessary to consider the synergistic influence of pore connectivity and sound field effects between voxel units in three-dimensional space. A hybrid architecture combining a three-dimensional convolutional neural network and a fully connected network is adopted. The input layer dimension matches the dimension of the pore feature tensor used for training, receiving high-dimensional tensor data of M×N×K×C, where M, N, and K are the number of voxels in three-dimensional space, and C is the number of pore feature parameter types. The three-dimensional convolution module contains four three-dimensional convolutional layers and ReLU activation function layers. The convolution kernel size is set to 3×3×3, progressively extracting the local and global distribution features of pore features in three-dimensional space. The ReLU activation function is used to introduce nonlinear transformations to enhance the model's expressive power. The pooling module uses max pooling layers with a pooling kernel size of 2×2×2 to downsample the feature map output by convolution, reducing the feature dimension while retaining key feature information. The fully connected module comprises two fully connected layers. The pooled 3D features are flattened into 1D feature vectors, which are then mapped to high-dimensional abstract features. The output layer dimension matches the parameter dimension of the training sound field control program, outputting the corresponding scanning path coordinate sequence, energy modulation parameters, and mode identifier. Furthermore, the output is not only a discrete set of parameters but also implicitly defines the dynamic scanning strategy and energy control logic of the sound field focal array in 3D space, thus constituting a set of control strategies for driving the growth of the porous structure.

[0034] Subsequently, an iterative training process is executed. The weight parameters of the network model are initialized using the He initialization method, assigning values ​​to the weights of each convolutional and fully connected layer, while the bias parameters are initialized to zero. The preprocessed training set samples are input into the model in batches of 16, and forward propagation is performed. The training aperture feature tensor is passed sequentially through the input layer, 3D convolutional module, pooling module, fully connected module, and output layer to obtain the predicted sound field control parameters. The mean squared error loss between the predicted parameters and the pre-calibrated training sound field control parameters is calculated and used as the objective function for model optimization. Backpropagation is then performed based on the loss value, using the Adam optimizer to calculate the gradients of the parameters in each layer. The network weights and bias parameters are updated with an initial learning rate of 0.001, and the number of iterations is set to 1000. After each iteration, the model performance is evaluated on the validation set, and the validation set loss value is calculated. When the validation set loss value no longer decreases after 20 consecutive iterations, or when the number of iterations reaches the preset upper limit, training is stopped and the current model parameters are saved, resulting in a trained process parameter generation network. This network can establish a high-dimensional nonlinear mapping relationship from complex three-dimensional pore configurations to dynamic sound field programming strategies.

[0035] Furthermore, the method provided in this application embodiment includes: The pore feature tensor is input into the process parameter generation network, and an inverse mapping operation is performed to generate a set of meta-instructions that define the structure growth strategy. The meta-instruction set specifies the phased control of the scanning path, energy temporal control waveform, and mode switching logic of the acoustic field focal array in different regions within the three-dimensional powder bed, which can induce powder particles to complete a preset self-organizing behavior. The meta-instruction set is compiled into a spatiotemporal sequence of instructions that can be recognized and executed by the underlying device of the programmable focused ultrasonic field, thus constituting the acoustic field control program.

[0036] In this embodiment, the three-dimensional powder bed is a three-dimensional powder accumulation layer with a preset thickness and loose stacking state formed by uniformly spreading metal powder in an ultrasonic-assisted metal powder forming process. It serves as the basic forming carrier for inducing the self-organization of powder particles to form the target pore structure. The acoustic field focal array is a collection of multiple acoustic field focal points with dynamically adjustable positions and energy formed within the three-dimensional powder bed by the coordinated operation of multiple ultrasonic transducers of a programmable focused ultrasonic field. This array is used for precise non-contact control of the metal powder particles.

[0037] Specifically, the pore feature tensor, after quantization, is first input into the input layer of the pre-trained process parameter generation network. The network's built-in tensor dimension matching method performs dimension verification and format adaptation on the input pore feature tensor, ensuring that the input data format is completely consistent with the input data format used during network training. After adaptation, the network's forward inference operation is initiated, sequentially passing the pore feature tensor through the network's 3D convolution module, pooling module, and fully connected module to extract the 3D spatial pore distribution features and pore parameter requirements for each region contained in the pore feature tensor. Using the high-dimensional nonlinear mapping relationship trained by the network, the inverse mapping operation from pore features to sound field control strategy is performed. Finally, a complete set of meta-instructions is output through the network's output layer, the specific content of which will be explained in detail in subsequent steps.

[0038] Next, the generated meta-instruction set adopts a structured hierarchical data structure, using the spatial voxel region of the three-dimensional powder bed as the basic unit. Each unit corresponds to an independent set of control parameters, which includes the spatial coordinate sequence of the sound field focal array, the energy output amplitude range, the action time node, and the mode switching identifier. This fully defines the sound field focal array scanning path, energy temporal control waveform, and mode switching logic for different regions and different forming stages within the three-dimensional powder bed. The meta-instruction set is located between the high-level pore structure representation and the low-level device control, serving as an intermediate control layer. It is used to transform the pore structure requirements in continuous space into discrete, executable sound field control strategies, unlike the traditional method of directly generating device control parameters. Through the non-contact action of the sound field, it can induce metal powder particles to complete self-organizing behavior that matches the target pore structure, providing a standardized strategy basis for the generation of subsequent sound field control programs.

[0039] Then, input the generated set of meta-instructions into the instruction compilation module supporting the programmable focused ultrasound field device. Using a general structured instruction compilation method, first perform a hierarchical analysis on the set of meta-instructions, split all the control parameter groups in the set of meta-instructions according to spatial regions and time stages, and map the abstract strategy parameters such as the spatial coordinate sequence, energy amplitude, action time, etc. in the split parameter groups into numerical control parameters recognizable by the underlying controller of the ultrasound field device. After completing the parameter mapping, sort the mapped control parameters in chronological order according to the time axis of device execution, generate single-step control instructions that are executed sequentially at time nodes, and then encapsulate all the single-step control instructions in the format according to the preset device communication protocol to form spatio-temporal sequence instructions recognizable and stably executed by the underlying device of the programmable focused ultrasound field. The spatio-temporal sequence instructions achieve coordinated control in the three-dimensional spatial dimension and the time dimension, enabling the sound field focus array to dynamically change along a preset evolution path in a continuous spatial region, thereby realizing the spatio-temporal coupling control of the pore structure formation process.

[0040] Through the inverse mapping operation of the pre-trained neural network, the automatic conversion from the target pore characteristics to the sound field control strategy is completed, and then the sound field control program directly executable by the device is generated through the structured instruction compilation method, realizing the accurate and efficient generation from the three-dimensional pore structure digital model to the sound field control program, providing a reliable control basis for the precise forming of the porous layer of metal powder in the follow-up.

[0041] Furthermore, the method provided in the embodiment of the present application includes: Path guiding strategy: To form a connected pore channel with a specific orientation, define that the sound field focus scans along a preset spatial curve and applies a periodically varying energy gradient; regionalized programming strategy: To achieve a non-monotonic spatial change in the porosity, divide the powder layer into multiple virtual voxel regions and independently define different sound field action programs; chronological modulation strategy: To optimize the pore morphology, define that the sound field energy and mode are dynamically modulated according to a preset time function at specific coordinate points.

[0042] Specifically, for connected pore channels with a specific orientation in the target pore structure, the spatial coordinates of the central axis of the target connected pore channels are first extracted from a pre-defined three-dimensional pore structure model. A three-dimensional curve fitting method is then used to fit the discrete coordinate points into a continuous and smooth pre-defined spatial curve, determining the start point, end point, and spatial coordinates and orientation of each node of the curve. Based on the pore diameter requirements of the pore channels, the scanning step size of the sound field focus is determined, set to 1 / 2 to 1 / 3 of the target pore diameter to ensure the continuity of the scanning trajectory. Along the fitted pre-defined spatial curve, the scanning path of the sound field focus is defined. Simultaneously, according to the porosity requirements at different locations in the pore channels, a linear gradient modulation method is used to set a periodically changing energy gradient. The highest energy amplitude is set at the central axis of the pore channel, and the energy amplitude is gradually linearly decreased towards both sides of the channel, forming a periodic energy distribution interval. Each energy cycle corresponds to the forming requirements of a segment of the pore channel. When the sound field focus scans along a preset spatial curve, the periodically changing energy gradient can induce the metal powder particles to directionally aggregate and rearrange along the scanning path, forming a continuous connected pore channel consistent with the preset direction. This completes the definition and integration of the path guidance strategy and writes it into the meta-instruction set.

[0043] To address the non-monotonic spatial variation requirements of porosity in the target pore structure, a three-dimensional spatial meshing method is employed. The entire three-dimensional space of the metal powder layer is divided into multiple independent virtual voxel regions. Each virtual voxel region corresponds to a unique spatial coordinate range. The region division size is determined based on the required accuracy of porosity variation, typically set to match the voxel unit size of the pore feature tensor to ensure the matching degree between the region division and the target pore features. For each virtual voxel region, pore feature parameters such as target porosity and average pore diameter are extracted from the corresponding region in the preset three-dimensional pore structure model. Corresponding acoustic field parameters, including acoustic field application time, energy amplitude, and focal point distribution density, are independently matched to each virtual voxel region. The parameters of different regions are independent and not forcibly correlated, enabling non-monotonic spatial variations such as jumps or gradual changes in porosity between adjacent regions. The acoustic field parameters corresponding to each virtual voxel region are integrated into independent acoustic field subroutines in spatial coordinate order. All subroutines together constitute the complete content of the regional programming strategy and are integrated into the meta-instruction set. They can then be directly compiled into device execution instructions for the corresponding region to achieve differentiated and independent control of the porosity characteristics of different regions.

[0044] For specific coordinate points in the target porous structure where pore morphology needs optimization, target pore morphology parameters corresponding to these coordinate points are extracted from the pore feature tensor, including pore roundness, pore wall roughness, and pore opening size. A function fitting method is used to generate a preset time function that modulates the sound field energy and mode over time based on the target pore morphology parameters. The time function can be a sine function, square wave function, or piecewise linear function, with its period, amplitude, and change nodes corresponding one-to-one with the target pore morphology parameters. In the meta-instruction set, the fitted preset time function is bound to this specific coordinate point. When the sound field acts at this coordinate point, the energy output amplitude and sound field focusing mode are dynamically modulated strictly according to the preset time function. Different energies and modes are output at different stages of pore formation. In the early stage of pore formation, a high-energy short-pulse mode is used to complete the initial arrangement of powder particles, while in the later stage, a low-energy continuous mode is used to optimize the pore wall morphology. This achieves precise control of the pore morphology at a specific coordinate point, completing the definition and implementation of the timing modulation strategy.

[0045] By employing three independent or combined acoustic field control strategies—path guidance, regional programming, and temporal modulation—precise and independent multi-dimensional control of the pore channel orientation, porosity spatial distribution, and pore morphology of porous metal powder layers is achieved. These strategies work synergistically within a unified set of meta-instructions. Through multi-dimensional coupling of spatial path control, regional parameter allocation, and dynamic temporal modulation, joint control of the pore structure formation process is realized, with different strategies mutually adaptable and synergistically optimized. This multi-strategy joint control method can achieve three-dimensional continuous interconnected pore structures and complex spatial gradient distributions that are difficult to obtain using traditional powder-laying-sintering processes. It is particularly suitable for the precise construction of complex pore channel structures such as spirals and twists required for high-flux tubes with twisted sintered porous layers, significantly improving the design freedom and performance control capabilities of porous layer structures, and providing core support for optimizing the heat transfer performance and operating condition adaptability of high-flux tubes.

[0046] Furthermore, the method provided in this application embodiment includes: The spatiotemporal sequence command of the programmable focused ultrasonic field can drive the ultrasonic field to form a focal array with a preset energy gradient in the powder layer. Through non-contact acoustic force and local thermal effect, it can actively guide the migration, rearrangement and self-assembly of powder particles, thereby programming and printing a porous precursor structure with preset three-dimensional connectivity and functional gradient.

[0047] Specifically, first, the spatio-temporal sequence instructions of the programmable focused ultrasound field after compilation are input into the industrial drive controller supporting the phased array ultrasonic transducer array. The drive controller adopts the general timing synchronization control method in the industrial control field and assigns corresponding emission phases, output amplitudes, and trigger times to each independent ultrasonic transducer unit in the array according to the time nodes and parameter requirements in the spatio-temporal sequence instructions. Each ultrasonic transducer unit emits ultrasonic waves synchronously according to the assigned parameters. Multiple ultrasonic waves undergo coherent superposition in the internal space of the three-dimensional powder bed, forming stable sound field focus points at the spatial coordinate positions preset by the spatio-temporal sequence instructions. All the sound field focus points are combined according to the preset spatial arrangement rules to form a complete sound field focus array. The drive controller forms a sound pressure distribution that gradually attenuates from the center of the focus point to the surrounding in the three-dimensional space covered by the sound field focus array by adjusting the output amplitudes, emission durations, and phase differences between adjacent transducers corresponding to different sound field focus points, and constructs a preset energy gradient matching the target pore structure.

[0048] Then, when the ultrasonic waves propagate in the three-dimensional powder bed, they continuously interact with the metal powder particles. Based on the basic principle of the acoustic radiation force, the sound waves are reflected, refracted, and scattered at the interface between the particles and the surrounding air medium, transferring the momentum they carry to the metal powder particles and generating an acoustic radiation force corresponding to the sound field energy distribution on the particles. The magnitude of the acoustic radiation force is positively correlated with the sound field energy gradient, and the direction is distributed along the direction of the energy gradient change. At the same time, part of the ultrasonic energy is absorbed by the metal powder particles and converted into heat energy through the acoustic-thermal conversion effect, resulting in an instantaneous temperature increase in the local area corresponding to the sound field focus point and generating a local thermal effect matching the sound field energy. The local thermal effect can reduce the contact friction resistance between the metal powder particles, improve the flow performance of the powder particles, and provide auxiliary conditions for the directional movement of the particles.

[0049] The preset energy gradient formed by the sound field focal array constructs a directional acoustic radiation force field within the powder layer. Driven by the acoustic radiation force, the metal powder particles migrate directionally along the energy gradient direction, moving towards the preset aggregation region with higher sound field energy, gradually completing the rearrangement and accumulation of particles. The spatial distribution and energy gradient of the aforementioned sound field focal array are not preset fixed patterns, but are obtained through reverse calculation using a process parameter generation network based on the target pore structure, ensuring a one-to-one correspondence between the sound field effect and the target pore structure in terms of spatial location and functional requirements. For high-density regions requiring the formation of a pore wall skeleton, high-energy sound field focal points are set through spatiotemporal sequence commands, guiding surrounding metal powder particles to continuously aggregate and densely accumulate in this region, forming the main skeleton of the pore structure. For channel regions requiring the formation of interconnected pores, low-energy sound field intervals are set through spatiotemporal sequence commands, causing powder particles to leave this region under the action of acoustic radiation force, forming continuous pore spaces. The acoustic field focal array dynamically evolves according to spatiotemporal sequence commands during execution. Its spatial position, energy distribution, and mode of action change continuously over time, thereby achieving dynamic control over the entire process of powder particle migration, rearrangement, and structure generation. Combined with the dynamic temporal control of the spatiotemporal sequence commands, the spatial position and energy distribution of the acoustic field focal array are gradually adjusted over time according to preset rules, continuously guiding the powder particles to complete self-assembly according to the requirements of the target pore structure, and gradually constructing a pore structure with three-dimensional connectivity.

[0050] Finally, by continuously executing complete spatiotemporal sequence commands through the drive controller, the sound field focus array completes the full-area coverage control of the entire three-dimensional space of the metal powder layer according to the preset scanning path and energy modulation rules, gradually completing the directional shaping of the entire pore structure, and finally obtaining a pore precursor structure with preset three-dimensional connectivity and functional gradient. The pore distribution, porosity and channel orientation of the precursor structure are completely matched with the target design requirements, and it can directly enter the subsequent sintering process to complete the final shaping of the twisted sintered porous high-flux tube.

[0051] By clarifying the execution logic of the spatiotemporal sequence command of the programmable focused ultrasound field, the formation mechanism of the sound field focal array and the preset energy gradient, and the principle of directional control of metal powder particles by acoustic radiation force and local thermal effect, the physical realization process of the porous precursor structure was reproduced, and non-contact precise programming control of the pore structure of the twisted sintered porous layer was realized.

[0052] Furthermore, the method provided in this application embodiment includes: The sound field control program is analyzed to extract the spatiotemporal sequence instructions and generate control signals for the powder spreading device and the programmable focused ultrasonic field that are strictly synchronized on the time axis. Based on the control signals for the powder spreading device, the powder spreading device is driven to spread metal powder on the surface of the spiral twisted base tube along a preset trajectory to form a powder prefabrication layer. Simultaneously, based on the control signals for the programmable focused ultrasonic field, the focal array of the programmable focused ultrasonic field is driven to dynamically track and act on a designated area of ​​the powder prefabrication layer in three-dimensional space, executing the scanning path, energy gradient, and mode switching in the spatiotemporal sequence instructions to programmatically guide the powder particles, thereby forming a pore precursor structure corresponding to the preset three-dimensional pore structure model in situ within the powder prefabrication layer.

[0053] Specifically, the sound field control program is first analyzed to extract the spatiotemporal sequence instructions. A structured instruction parsing method is used to split the spatial coordinate parameters, energy parameters, and mode parameters within the instructions according to the timestamp dimension. The split parameters are then mapped to powder spreading device control signals and programmable focused ultrasound field control signals, respectively. The two types of control signals are time-calibrated using a unified time axis reference to ensure that the execution nodes of the two are strictly synchronized on the time axis, and the synchronization error is controlled within a preset threshold.

[0054] Next, based on the generated powder spreading device control signal, the powder spreading device is driven by a servo drive control method. The control signal contains motion trajectory parameters, powder spreading speed parameters, and powder spreading thickness parameters, which guide the powder spreading device to move along a preset trajectory that matches the shape of the spiral twisted base tube, so as to evenly spread metal powder on the surface of the base tube and gradually form a continuous powder prefabricated layer. During the powder spreading process, the powder spreading speed and powder spreading thickness are kept stable to ensure the uniformity and integrity of the powder prefabricated layer.

[0055] While the powder-laying device performs its laying action, a phased array ultrasonic dynamic focusing control method is used based on the programmable focused ultrasonic field control signal to drive the ultrasonic transducer array to adjust the emission phase and output amplitude of each unit. This allows the acoustic field focus array to dynamically track the currently laid powder prefabricated layer area in three-dimensional space, move the focus according to the scanning path in the spatiotemporal sequence command, adjust the output energy of each focus according to the preset energy gradient, and switch the acoustic field mode at specified time nodes. The motion trajectory of the powder-laying device and the scanning path of the acoustic field focus array are coupled based on the same spatiotemporal sequence command, ensuring that the acoustic field action area always coincides with the spatial area of ​​the currently laid powder, thereby achieving synchronous coupling between the powder-laying process and the pore structure construction process.

[0056] By guiding the directional migration and rearrangement of powder particles through acoustic force and local thermal effect, a pore precursor structure corresponding to the preset three-dimensional pore structure model is formed in situ within the powder prefabrication layer. The pore precursor structure is constructed in real time during the powder laying process, without relying on the passive pore generation process in the subsequent sintering stage, thereby realizing active pore structure forming.

[0057] Furthermore, the method provided in this application embodiment includes: While the programmable focused ultrasonic field irradiates and modulates the powder prefabricated layer, an integrated acoustic sensor collects acoustic feedback signals generated by the scattering of powder particles in real time. Based on a pre-trained acoustic feature analysis model, the acoustic feedback signals are processed to interpret the changes in powder accumulation state and the progress of pore structure formation in the currently irradiated and modulated area in real time. The interpreted changes in powder accumulation state and the progress of pore structure formation are compared with the expected state of the currently executed meta-instruction set. If a deviation occurs, the parameters of the subsequent meta-instruction set that has not yet been executed are dynamically adjusted to achieve adaptive optimization of the strategy for the structure growth process.

[0058] In one embodiment, firstly, multiple broadband piezoacoustic sensors are integrated and arranged around the phased array transducer array of the programmable focused ultrasonic field. The sensors and transducer array are rigidly fixed, and the acquisition frequency range covers 1 to 3 times the main ultrasonic emission frequency to avoid signal aliasing. While the programmable focused ultrasonic field irradiates and modulates the powder prefabricated layer, a multi-channel synchronous data acquisition method is used. An industrial-grade data acquisition card is used to synchronously acquire the output signals of all acoustic sensors at a preset sampling frequency. The sampling frequency is set to more than 10 times the ultrasonic emission frequency to ensure the integrity of the acquired signals. The acquired signal is the acoustic feedback signal generated by the scattering and reflection of the incident ultrasonic waves by the powder particles. The acquisition process is strictly time-synchronized with the ultrasonic field irradiation and modulation process, with the synchronization error controlled within 1 microsecond. The acquired acoustic feedback signal undergoes conventional bandpass filtering to remove environmental noise and low-frequency mechanical vibration interference, resulting in a pre-processed effective acoustic feedback signal.

[0059] Then, an acoustic feature analysis model was built using a one-dimensional convolutional neural network architecture. The model consists of an input layer, three cascaded one-dimensional convolutional modules, a pooling module, a fully connected layer, and an output layer. Each one-dimensional convolutional module contains 16 convolutional kernels with a kernel size of 3. The ReLU activation function is used, and the pooling module uses max pooling with a pooling window size of 2. The model input is a preprocessed time series array of one-dimensional acoustic feedback signals, and the model output consists of three quantitative parameters: pore structure formation degree, particle packing density, and pore defect risk value of the current printing area, corresponding to the pore structure formation progress, powder packing state change, and potential defect initiation signal, respectively.

[0060] Subsequently, a training dataset for the model was constructed using the controlled variable method. Under powder spreading conditions and ultrasonic field parameters consistent with the actual molding process, standard samples with different pore structure formation degrees, particle packing densities, and defect types were prepared. Acoustic feedback signals were simultaneously acquired for each sample during preparation. Industrial CT scanning was used to obtain the actual labeled values ​​of pore structure formation degree, particle packing density, and defect state in the corresponding region of each sample. At least 100 valid samples were prepared for each parameter gradient. Finally, a training dataset containing at least 5000 samples and a validation dataset containing at least 1000 samples were constructed. Supervised learning was used to train the model, using the acoustic feedback signals from the training dataset as input and the corresponding labeled values ​​as labels. The mean squared error function was used as the loss function, and the Adam optimizer was used for iterative optimization of the model parameters. The training batch size was set to 32, and the training iterations were set to 100 rounds. After each training round, the model accuracy was verified using the validation dataset. Training was stopped when the prediction error on the validation dataset was less than 5%, resulting in the pre-trained acoustic feature analysis model.

[0061] Next, the preprocessed effective acoustic feedback signal is input into the input layer of the pre-trained acoustic feature analysis model. The model's one-dimensional convolutional module automatically extracts the temporal, frequency, and energy distribution features from the acoustic feedback signal. The extracted features include peak signal amplitude, dominant frequency offset, scattered signal attenuation coefficient, and harmonic energy ratio. These features have a direct correspondence with the powder packing state and pore structure formation state. After feature extraction, feature mapping and quantization calculations are performed through a fully connected layer. Finally, the output layer outputs the pore structure formation degree, particle packing density, and pore defect risk value of the currently irradiated and modulated area. The changes in powder packing state and pore structure formation progress in the area are interpreted in real time. The single computation time of the interpretation process does not exceed 10 milliseconds, meeting the requirements of online real-time processing.

[0062] Finally, the changes in powder accumulation state and pore structure formation progress in the current region, obtained from the model interpretation above, are compared in real time with the expected state parameters of the corresponding region and time node in the currently executed meta-instruction set to calculate the deviation between the actual and expected parameters. When the deviation exceeds the preset allowable deviation threshold, an adaptive adjustment process is triggered. Using a proportional-integral-derivative control method, the corresponding parameters in the subsequent unexecuted meta-instruction set are dynamically adjusted based on the magnitude and trend of the deviation. The adjusted parameters include the sound field focus scanning step size, energy gradient amplitude, duration of action, and mode switching node. This adaptive adjustment does not directly modify the underlying device control parameters but acts on the meta-instruction set layer, updating the structure growth strategy to achieve overall optimization of subsequent sound field control behavior. The adjusted meta-instruction set is recompiled into a spatiotemporal sequence of instructions for the subsequent sound field control program, achieving online adaptive optimization of the pore structure growth process. The pore defect risk value obtained in the above steps is used to predict the formation trend of potential defects, enabling the control system to adjust the sound field control strategy in advance before defects actually form, achieving predictive control.

[0063] By synchronously acquiring acoustic feedback signals during the molding process, combining them with a pre-trained acoustic feature analysis model to interpret the molding state of the pore structure in real time, and dynamically adjusting the meta-instruction set parameters based on the state deviation, online adaptive closed-loop control of the molding process of the twisted sintered porous high-flux tube was realized. This effectively reduced the molding deviation and defect risk, and improved the molding accuracy and batch consistency of the pore structure.

[0064] Furthermore, the method provided in this application embodiment includes: The sound field control program is input into a high-fidelity process simulation model to simulate the complete irradiation modulation process defined by the sound field control program and predict the resulting simulated pore structure. Simulated pore features are extracted from the simulated pore structure, and the predicted structural deviation between the simulated pore features and the target pore features represented by the pore feature tensor is calculated. It is determined whether the predicted structural deviation exceeds the verification threshold. If it does, the predicted structural deviation and the corresponding simulated pore structure are used as feedback to re-trigger the process parameter generation network to reverse the mapping of the pore feature tensor, generating an updated sound field control program until the predicted structural deviation meets the verification threshold.

[0065] Optionally, the generated acoustic field control program is first input into a pre-built high-fidelity process simulation model. This simulation model can be built using EDEM discrete element software and COMSOL Multiphysics software, achieving bidirectional coupling between the discrete element field and the acoustic physical field through a real-time data interaction interface between the software. It integrates the mechanical properties of metal powder particles, the law of acoustic radiation force, and the acoustic-thermal conversion effect, enabling a complete reproduction of the entire process from powder spreading motion, acoustic field irradiation modulation, powder particle migration and rearrangement, to pore structure formation. After starting the simulation model, the coordinated action of the powder spreading device movement and the programmable focused ultrasonic field output is simulated stage by stage according to the spatiotemporal sequence instructions defined in the acoustic field control program. This fully reproduces all process steps from powder prefabrication layer laying to the formation of the pore precursor structure, ultimately outputting a simulated pore structure corresponding to the actual forming result.

[0066] Then, feature extraction is performed on the simulated pore structure output from the simulation. Using 3D image analysis, parameters such as porosity, 3D connectivity, pore orientation distribution, pore size distribution, and pore morphology are extracted to form a simulated pore feature set with the same dimension as the target pore feature tensor. Next, min-max normalization is performed on each type of feature parameter in both the simulated pore feature set and the target pore feature tensor, linearly mapping the parameter values ​​to the 0-1 range to eliminate dimensional differences between different parameters. Based on the normalized features, the absolute difference between the simulated and target values ​​for each type of feature is calculated. Preset weights are then assigned according to the influence of each type of feature on the molding accuracy. The weighted differences are summed to obtain the predictive structural deviation between the two feature sets. This deviation is quantified into a directly comparable numerical index between 0 and 1.

[0067] Finally, the calculated predicted structural deviation is compared with a preset verification threshold. The verification threshold is preset according to the product molding accuracy requirements. For the predicted structural deviation obtained in the above steps, 0 represents a complete match and 1 represents a complete mismatch. The verification threshold can be directly set to 0.05 or 0.1 for easy engineering judgment. If the predicted structural deviation does not exceed the verification threshold, the current sound field control program is determined to meet the process requirements, and the actual irradiation modulation execution stage can be entered. If the predicted structural deviation exceeds the verification threshold, the deviation value and the corresponding simulated pore structure data are fed back to the process parameter generation network, triggering the network to perform inverse mapping calculation of the pore feature tensor, updating the meta-instruction set parameters in the sound field control program, including the sound field focus scanning path, energy gradient amplitude, and timing modulation rules, generating an updated sound field control program. The updated sound field control program is then input into the high-fidelity process simulation model again, and the above simulation, feature extraction, deviation calculation, and comparison process is repeated until the predicted structural deviation meets the verification threshold requirements. The high-fidelity process simulation model and the process parameter generation network form a coupled computational system. The simulation results are used to participate in the inverse mapping process to achieve iterative optimization of the sound field control strategy.

[0068] Virtual pre-verification using a high-fidelity process simulation model allows for the verification and iterative optimization of the sound field control program before actual process execution. This avoids trial-and-error costs in the actual molding process, improves the molding accuracy and process stability of the pore structure, and enables the prediction of pore structure molding results without actual physical experiments, significantly reducing the number of trial-and-error attempts during process debugging.

[0069] Furthermore, the method provided in this application embodiment includes: The sound field control program is input into a high-fidelity process simulation model to simulate the complete irradiation modulation process defined by the sound field control program and predict the resulting simulated pore structure. Simulated pore features are extracted from the simulated pore structure, and the predicted structural deviation between the simulated pore features and the target pore features represented by the pore feature tensor is calculated. It is determined whether the predicted structural deviation exceeds the verification threshold. If it does, the predicted structural deviation and the corresponding simulated pore structure are used as feedback to re-trigger the process parameter generation network to reverse the mapping of the pore feature tensor, generating an updated sound field control program until the predicted structural deviation meets the verification threshold.

[0070] In one embodiment, the actual molded product obtained after sintering is first subjected to non-destructive testing. An industrial computer tomography (CT) scanner is used, and matching scanning resolution and scanning parameters are set to acquire three-dimensional voxel image data of the porous layer of metal powder. Through three-dimensional image analysis and processing, parameters such as porosity, three-dimensional connectivity, channel orientation distribution, pore size distribution, and pore morphology of the actual pore structure are extracted to obtain actual pore structure data reflecting the actual molding state.

[0071] Next, the extracted actual pore structure data is quantized into an actual pore feature tensor according to the same dimensions and format as the target pore feature tensor. For each type of feature parameter in both the actual and target pore feature tensors, the same min-max normalization process as described above is performed, linearly mapping the parameter values ​​to the 0-1 interval to eliminate dimensional differences between different parameters. The absolute difference between the actual and target values ​​of each feature type is calculated, and preset weights are assigned according to the degree of influence of each feature on product performance. The weighted differences are then summed to obtain the actual structural deviation between the two feature tensors.

[0072] Then, the calculated actual structural deviation, the corresponding target pore feature tensor, and the acoustic field control program used in this molding process are paired to form a complete set of incremental training samples. These incremental training samples are input into the pre-trained process parameter generation network. Using a small learning rate incremental learning strategy, with the actual structural deviation as the optimization target, the parameters of the fully connected layers and some convolutional layers of the network are fine-tuned and iteratively updated. This retains the network's original inverse mapping capability while adapting to the feature distribution of the actual process conditions, gradually improving the network's generation accuracy of the acoustic field control program. The acquisition and feedback of the actual structural deviation achieves a cross-stage data loop from the actual molding process to the process parameter generation network, enabling a continuously interactive optimization system between the physical manufacturing process and the computational model.

[0073] Through a complete process of sintering, non-destructive testing, deviation quantification, and incremental learning, the process parameter generation network was continuously iteratively optimized. This provides a more precise acoustic field control program for subsequent molding processes, further improving the molding accuracy and batch stability of the product's pore structure. The incremental learning process can gradually adapt to changes in the material characteristics and equipment operating conditions of different batches, improving the robustness and generalization ability of the process parameter generation network in actual production environments.

[0074] In summary, the optimized forming process method for the twisted sintered porous layer high-flux tube provided in this application embodiment has the following technical effects: This application generates a sound field control program and simultaneously coordinates powder spreading and focused ultrasonic field irradiation. Through acoustic feedback analysis, virtual simulation verification, and sintering and forming detection, it calculates pore structure deviations and iteratively optimizes the sound field control program. Combined with actual forming data, it fine-tunes process parameters to generate a network, thereby accurately forming a three-dimensional pore structure. This significantly improves the forming accuracy and detection reliability of the twisted sintered porous layer high-throughput tube, achieving the technical effect of precise forming of porous layer pore structure and dynamic control of the entire process.

[0075] Example 2, as Figure 2 As shown, based on the same inventive concept as in Embodiment 1 above, this application provides a forming process optimization system for a twisted sintered porous layer high-flux tube, the system comprising: The pore feature tensor acquisition module 1 is used to acquire a preset three-dimensional pore structure model of the target product. The preset three-dimensional pore structure model defines the target pore features of the metal powder porous layer of the high-throughput tube at different spatial locations using a digital model, and quantifies the target pore features into a computable pore feature tensor.

[0076] The sound field control program output module 2 is used to input the pore feature tensor into a pre-trained process parameter generation network, reverse map and output a sound field control program that can realize the target pore features, wherein the sound field control program includes spatiotemporal sequence instructions for controlling the programmable focused ultrasonic field.

[0077] The porous precursor structure construction module 3 is used to synchronously coordinate and control the motion parameters of the powder laying device and the output parameters of the programmable focused ultrasonic field according to the spatiotemporal sequence instructions in the acoustic field control program during the metal powder laying process, so that the programmable focused ultrasonic field performs real-time irradiation modulation on the powder prefabrication layer according to the preset program, so as to form a porous precursor structure in the powder prefabrication layer.

[0078] The process parameter generation network update module 4 is used to sinter the powder preform layer with porous precursor structure and the spiral twisted base tube, and after sintering, collect the real-time pore structure data of the actual molded product, calculate the structural deviation with the pore feature tensor, and iteratively update the process parameter generation network based on the calculated structural deviation.

[0079] Furthermore, the porous precursor structure construction module 3 is used to perform the following steps: The spatiotemporal sequence command of the programmable focused ultrasonic field can drive the ultrasonic field to form a focal array with a preset energy gradient in the powder layer. Through non-contact acoustic force and local thermal effect, it can actively guide the migration, rearrangement and self-assembly of powder particles, thereby programming and printing a porous precursor structure with preset three-dimensional connectivity and functional gradient.

[0080] Furthermore, the sound field control program output module 2 is used to perform the following steps: A training dataset containing multiple sets of sample pairs is constructed, wherein each set of sample pairs contains a training pore feature tensor as the network input and a pre-calibrated training sound field control program as the expected output of the network; the training pore feature tensor is used as the input and the training sound field control program is used as the expected output, and the neural network architecture is used for iterative training until a high-dimensional nonlinear mapping relationship is established between the complex three-dimensional pore configuration and the dynamic sound field programming strategy required to realize the configuration, thereby generating a process parameter generation network.

[0081] Furthermore, the sound field control program output module 2 is used to perform the following steps: The pore feature tensor is input into the process parameter generation network, and an inverse mapping operation is performed to generate a set of meta-instructions that define the structure growth strategy. The meta-instruction set specifies the phased control of the scanning path, energy temporal control waveform, and mode switching logic of the acoustic field focal array in different regions within the three-dimensional powder bed, which can induce powder particles to complete a preset self-organizing behavior. The meta-instruction set is compiled into a spatiotemporal sequence of instructions that can be recognized and executed by the underlying device of the programmable focused ultrasonic field, thus constituting the acoustic field control program.

[0082] Furthermore, the sound field control program output module 2 is used to perform the following steps: The path-guided strategy defines the sound field focus to scan along a preset spatial curve and apply a periodically changing energy gradient in order to form a connected pore channel with a specific orientation. The regional programming strategy divides the powder layer into multiple virtual voxel regions and independently defines differentiated sound field action programs to achieve non-monotonic spatial variation of porosity. The temporal modulation strategy defines the sound field energy and mode at specific coordinate points and dynamically modulates them according to a preset time function to optimize pore morphology.

[0083] Furthermore, the porous precursor structure construction module 3 is used to perform the following steps: The sound field control program is analyzed to extract the spatiotemporal sequence instructions and generate control signals for the powder spreading device and the programmable focused ultrasonic field that are strictly synchronized on the time axis. Based on the control signals for the powder spreading device, the powder spreading device is driven to spread metal powder on the surface of the spiral twisted base tube along a preset trajectory to form a powder prefabrication layer. Simultaneously, based on the control signals for the programmable focused ultrasonic field, the focal array of the programmable focused ultrasonic field is driven to dynamically track and act on a designated area of ​​the powder prefabrication layer in three-dimensional space, executing the scanning path, energy gradient, and mode switching in the spatiotemporal sequence instructions to programmatically guide the powder particles, thereby forming a pore precursor structure corresponding to the preset three-dimensional pore structure model in situ within the powder prefabrication layer.

[0084] Furthermore, the porous precursor structure construction module 3 is used to perform the following steps: While the programmable focused ultrasonic field irradiates and modulates the powder prefabricated layer, an integrated acoustic sensor collects acoustic feedback signals generated by the scattering of powder particles in real time. Based on a pre-trained acoustic feature analysis model, the acoustic feedback signals are processed to interpret the changes in powder accumulation state and the progress of pore structure formation in the currently irradiated and modulated area in real time. The interpreted changes in powder accumulation state and the progress of pore structure formation are compared with the expected state of the currently executed meta-instruction set. If a deviation occurs, the parameters of the subsequent meta-instruction set that has not yet been executed are dynamically adjusted to achieve adaptive optimization of the strategy for the structure growth process.

[0085] Furthermore, the porous precursor structure construction module 3 is used to perform the following steps: The sound field control program is input into a high-fidelity process simulation model to simulate the complete irradiation modulation process defined by the sound field control program and predict the resulting simulated pore structure. Simulated pore features are extracted from the simulated pore structure, and the predicted structural deviation between the simulated pore features and the target pore features represented by the pore feature tensor is calculated. It is determined whether the predicted structural deviation exceeds the verification threshold. If it does, the predicted structural deviation and the corresponding simulated pore structure are used as feedback to re-trigger the process parameter generation network to reverse the mapping of the pore feature tensor, generating an updated sound field control program until the predicted structural deviation meets the verification threshold.

[0086] Furthermore, the process parameter generation network update module 4 is used to perform the following steps: Non-destructive testing is performed on the actual sintered product to obtain three-dimensional image data of its porous metal powder layer, and actual pore structure data is extracted from it. The actual pore structure data is quantized into an actual pore feature tensor, and the actual structural deviation between the actual pore feature tensor and the pore feature tensor is calculated. The actual structural deviation, the pore feature tensor, and the actual sound field control program are used together to form a set of incremental training samples. The incremental training samples are used to fine-tune and iteratively update the process parameter generation network.

[0087] The forming process optimization system for the twisted sintered porous layer high-throughput tube provided in this embodiment of the invention can execute the forming process optimization method for the twisted sintered porous layer high-throughput tube provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0088] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0089] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. An optimized method for the forming process of a twisted sintered porous layer high-flux tube, characterized in that, The method includes: A preset three-dimensional pore structure model of the target product is obtained. The preset three-dimensional pore structure model defines the target pore features of the metal powder porous layer of the high-throughput tube at different spatial locations using a digital model, and quantifies the target pore features into a computable pore feature tensor. The pore feature tensor is input into a pre-trained process parameter generation network, which is then inversely mapped and outputs an acoustic field control program that can achieve the target pore features. The acoustic field control program includes a spatiotemporal sequence instruction for controlling the programmable focused ultrasonic field. During the metal powder laying process, according to the spatiotemporal sequence instructions in the sound field control program, the motion parameters of the powder laying device and the output parameters of the programmable focused ultrasonic field are synchronously coordinated and controlled, so that the programmable focused ultrasonic field performs real-time irradiation modulation on the powder preform layer according to the preset program, so as to form a porous precursor structure in the powder preform layer. The powder preform layer with a porous precursor structure and the spiral twisted base tube are sintered. After sintering, real-time pore structure data of the actual molded product are collected, and structural deviation is calculated with the pore feature tensor. The process parameter generation network is iteratively updated based on the calculated structural deviation.

2. The optimized forming process method for the twisted sintered porous layer high-flux tube as described in claim 1, characterized in that, The spatiotemporal sequence command of the programmable focused ultrasonic field can drive the ultrasonic field to form a focal array with a preset energy gradient in the powder layer. Through non-contact acoustic force and local thermal effect, it can actively guide the migration, rearrangement and self-assembly of powder particles, thereby programming and printing a porous precursor structure with preset three-dimensional connectivity and functional gradient.

3. The optimized forming process method for the twisted sintered porous layer high-flux tube as described in claim 1, characterized in that, The training of the process parameter generation network includes: Construct a training dataset containing multiple sets of sample pairs, where each set of sample pairs contains a training aperture feature tensor as input to the network and a pre-calibrated training acoustic field modulation procedure as the expected output of the network. Using the training pore feature tensor as input and the training sound field control program as expected output, iterative training is performed in conjunction with a neural network architecture until a high-dimensional nonlinear mapping relationship is established between the complex three-dimensional pore configuration and the dynamic sound field programming strategy required to achieve the configuration, thereby generating a process parameter generation network.

4. The optimized forming process method for the twisted sintered porous layer high-flux tube as described in claim 3, characterized in that, The pore feature tensor is input into a pre-trained process parameter generation network, which is then inversely mapped and outputs an acoustic field control program that can achieve the target pore features. The acoustic field control program includes spatiotemporal sequence instructions for controlling the programmable focused ultrasonic field, including: The pore feature tensor is input into the process parameter generation network, and an inverse mapping operation is performed to generate a set of meta-instructions that define the structure growth strategy. The meta-instruction set specifies the scanning path, energy temporal control waveform and mode switching logic of the acoustic field focal array in different regions of the three-dimensional powder bed in stages, which can induce powder particles to complete the preset self-organization behavior. The set of meta-instructions is compiled into a spatiotemporal sequence of instructions that can be recognized and executed by the underlying device of the programmable focused ultrasound field, thus constituting the sound field control program.

5. The method for optimizing the forming process of the twisted sintered porous layer high-flux tube as described in claim 4, characterized in that, The meta-instruction set includes at least one of the following strategies: The path-guided strategy defines the sound field focus to scan along a preset spatial curve and applies a periodically changing energy gradient in order to form a connected pore channel with a specific orientation. The regional programming strategy divides the powder layer into multiple virtual voxel regions and independently defines differentiated acoustic field interaction programs to achieve non-monotonic spatial variation of porosity. The timing modulation strategy optimizes the pore morphology by dynamically modulating the sound field energy and mode at specific coordinate points according to a preset time function.

6. The optimized forming process method for the twisted sintered porous layer high-flux tube as described in claim 1, characterized in that, During the metal powder laying process, according to the spatiotemporal sequence instructions in the acoustic field control program, the motion parameters of the powder laying device and the output parameters of the programmable focused ultrasonic field are synchronously and coordinated to enable the programmable focused ultrasonic field to perform real-time irradiation modulation on the powder preform layer according to a preset program, so as to form a porous precursor structure within the powder preform layer, including: The sound field control program is analyzed, the spatiotemporal sequence instructions are extracted, and control signals for the powder spreading device and programmable focused ultrasonic field are generated that are strictly synchronized on the time axis. According to the control signal of the powder spreading device, the powder spreading device is driven to spread metal powder on the surface of the spiral twisted base tube along a preset trajectory to form a powder prefabrication layer; Synchronously, based on the programmable focused ultrasonic field control signal, the focal array of the programmable focused ultrasonic field is driven to dynamically track and act on a designated area of ​​the powder preform in three-dimensional space, execute the scanning path, energy gradient and mode switching in the spatiotemporal sequence command, and programmatically guide the powder particles, thereby forming a pore precursor structure corresponding to the preset three-dimensional pore structure model in situ within the powder preform.

7. The optimized forming process method for the twisted sintered porous layer high-flux tube as described in claim 6, characterized in that, The real-time irradiation modulation process also includes an online adaptive control step based on acoustic feedback, including: While the programmable focused ultrasonic field irradiates and modulates the powder preform, an integrated acoustic sensor collects the acoustic feedback signal generated by the scattering of powder particles in real time. Based on the pre-trained acoustic feature analysis model, the acoustic feedback signal is processed to interpret in real time the changes in powder accumulation state and pore structure formation progress in the irradiated modulation area. The decoded changes in powder accumulation state and pore structure formation progress are compared with the expected state of the currently executed meta-instruction set. If a deviation occurs, the parameters of the subsequent meta-instruction set that has not yet been executed are dynamically adjusted to achieve adaptive optimization of the strategy for the structure growth process.

8. The optimized forming process method for the twisted sintered porous layer high-flux tube as described in claim 7, characterized in that, Before executing the real-time irradiation modulation step according to the sound field control procedure, a virtual verification step is also included: The sound field control program is input into the high-fidelity process simulation model to simulate the complete irradiation modulation process defined by the sound field control program and predict the simulated pore structure to be obtained. Simulated pore features are extracted from the simulated pore structure, and the predictive structural deviation between the simulated pore features and the target pore features represented by the pore feature tensor is calculated. If the predicted structural deviation exceeds the verification threshold, the predicted structural deviation and the corresponding simulated pore structure are used as feedback to re-trigger the process parameter generation network to reverse the mapping of the pore feature tensor, generating an updated sound field control program until the predicted structural deviation meets the verification threshold.

9. The optimized forming process method for the twisted sintered porous layer high-flux tube as described in claim 8, characterized in that, After sintering, real-time pore structure data of the actual molded product is collected, and structural deviation is calculated between the data and the pore feature tensor. Based on the calculated structural deviation, the process parameter generation network is iteratively updated, including: Non-destructive testing was performed on the actual molded product obtained after sintering to obtain three-dimensional image data of its metal powder porous layer and extract the actual pore structure data from it. The actual pore structure data is quantized into an actual pore feature tensor, and the actual structural deviation between the actual pore feature tensor and the pore feature tensor is calculated. The actual structural deviation, the pore feature tensor, and the actual sound field control program used together constitute a set of incremental training samples. The incremental training samples are used to fine-tune and iteratively update the parameters of the process parameter generation network.

10. A molding process optimization system for twisted sintered porous layer high-flux tubes, characterized in that, A method for optimizing the forming process of the twisted sintered porous layer high-flux tube according to any one of claims 1-9, the system comprising: The pore feature tensor acquisition module is used to acquire a preset three-dimensional pore structure model of the target product. The preset three-dimensional pore structure model defines the target pore features of the metal powder porous layer of the high-throughput tube at different spatial locations using a digital model, and quantifies the target pore features into a computable pore feature tensor. The acoustic field control program output module is used to input the pore feature tensor into a pre-trained process parameter generation network, reverse map and output an acoustic field control program that can realize the target pore features, wherein the acoustic field control program includes spatiotemporal sequence instructions for controlling the programmable focused ultrasonic field. The porous precursor structure construction module is used to synchronously coordinate and control the motion parameters of the powder laying device and the output parameters of the programmable focused ultrasonic field according to the spatiotemporal sequence instructions in the acoustic field control program during the metal powder laying process, so that the programmable focused ultrasonic field can irradiate and modulate the powder prefabrication layer in real time according to the preset program to form a porous precursor structure in the powder prefabrication layer. The process parameter generation network update module is used to sinter the powder preform layer with porous precursor structure and the spiral twisted base tube. After sintering, it collects the real-time pore structure data of the actual molded product, calculates the structural deviation with the pore feature tensor, and iteratively updates the process parameter generation network based on the calculated structural deviation.

Images