Method for controlling wall thickness uniformity of additive manufacturing combustion channel of aerospace propulsion chamber

Abstract

The invention discloses a spaceflight propulsion chamber additive manufacturing combustion channel wall thickness uniformity control method, which comprises the following steps: S1, establishing a three-dimensional geometric model and carrying out layered slicing to generate an initial processing control parameter set; s2, constructing a thermal echo spectrum before printing the first layer, and correcting control parameters of the first layer; s3, generating a wall thickness error graph of the current layer based on the data of the previous layer, and constructing a wall thickness error potential energy field graph; s4, constructing a target scanning density function, solving a two-dimensional Monge-Amprore equation to generate a path potential function, and generating a quasi potential streamline field; s5, inputting the data of the current layer into the wall thickness prediction neural network model, and outputting a wall thickness deviation prediction value of the next layer; s6, updating the wall thickness error graph of the next layer according to the wall thickness deviation predicted value, and dynamically adjusting the control parameters; and S7, circularly executing the steps S3 to S6 until the printing task is completed. According to the method, thermal response modeling and a neural network model are fused, and dynamic uniformity control over the wall thickness of the combustion channel is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of additive manufacturing control, and particularly to a method for controlling the wall thickness uniformity of a combustion channel in the additive manufacturing of a space propulsion chamber. Background Art

[0002] With the continuous development of space propulsion technology, higher requirements are put forward for the structural complexity, precision, and stability of core thermal structure components such as engine combustion channels. Selective Laser Melting (SLM), as a high-precision and high-degree-of-freedom metal additive manufacturing technology, has become an important means for manufacturing complex internal flow channel structures such as combustion channels in space propulsion chambers. However, due to the characteristics of the combustion channel wall structure being limited by its complex curved surface, closed rotation, and intensive heat load, quality problems such as uneven wall thickness, geometric distortion, unstable molten pool, and stress concentration are extremely likely to occur in the actual additive manufacturing process, seriously affecting the subsequent processing matching and service safety of components.

[0003] Currently, the control of the wall thickness in additive manufacturing usually relies on preset scanning strategies and empirical process parameters. For example, serpentine filling, staggered scanning, and power adjustment methods are used to statically optimize the forming process. However, most of these control methods are based on global parameter settings, lacking fine modeling and dynamic response capabilities for the printing states at different positions, and it is difficult to achieve local prediction and closed-loop correction of wall thickness errors. In addition, traditional path planning methods mostly adopt straight line or uniform distribution strategies, failing to fully combine the changing characteristics of the spatial error distribution of the wall thickness for adaptive path reconstruction, resulting in redundant or insufficient accumulation in local areas, thereby causing the accumulation of wall thickness deviation.

[0004] In terms of measurement feedback, most of the existing technologies rely on linear profile measurement or surface point cloud reconstruction methods after printing to obtain structural deviation information, and it is impossible to achieve dynamic modeling and real-time predictive control during the printing process. At the same time, even in some studies, infrared thermal imaging or molten pool monitoring means are introduced, but due to the lack of systematic data modeling and physical interpretation paths, it is difficult to form an effective correlation mechanism between the thermal response behavior and the subsequent wall thickness error, and thus it is impossible to effectively convert the thermal process characteristics into control feedback basis.

[0005] In recent years, some studies have begun to attempt to model and predict the additive manufacturing process by combining deep learning methods. For example, convolutional neural networks are used to extract topography features or recurrent neural networks are introduced to analyze historical data trends. However, since most of these models lack physical guiding mechanisms and have a single input dimension, the prediction results are difficult to map to the specific process regulation parameter level. At the same time, there is currently a lack of dynamic control logic for two-dimensional spatial error fields, and it is impossible to establish a fine control strategy allocation mechanism for each scanning unit, resulting in a disconnection between prediction and adjustment and unable to form a complete closed-loop feedback path.

[0006] Therefore, how to provide a method for controlling the wall thickness uniformity of the combustion channel in the additive manufacturing of aerospace propulsion chambers is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0007] An object of the present invention is to propose a method for controlling the wall thickness uniformity of the combustion channel in the additive manufacturing of aerospace propulsion chambers. The present invention integrates thermal response modeling, the construction of a wall thickness error potential energy field, and a deep learning prediction network. It obtains thermal property differences through non-melting thermal response scanning, dynamically reconstructs the scanning path based on the Monge-Ampère equation, and combines a multi-dimensional wall thickness prediction model to achieve point-by-point adaptive adjustment of control parameters. It has the advantages of high wall thickness uniformity, fast error response, and high path control accuracy, effectively improving the forming quality and structural consistency of the combustion channel in the aerospace propulsion chamber during the additive manufacturing process, and solving the problems of path rigidity, regulation lag, and rough wall thickness control in the existing manufacturing process.

[0008] A method for controlling the wall thickness uniformity of the combustion channel in the additive manufacturing of aerospace propulsion chambers according to an embodiment of the present invention includes the following steps: S1. Establish a three-dimensional geometric model of the combustion channel and perform layer slicing to obtain the structural information of each layer, and generate an initial processing control parameter set; S2. Before printing the first layer, use a frequency-variable laser to perform non-melting scanning on the area of the combustion channel to be printed and construct a thermal echo map, and correct the control parameters for the first layer in the initial processing control parameter set according to the thermal echo map; S3. Before printing each layer, generate a wall thickness error map for the current layer based on the actual forming data of the previous layer, and map the wall thickness error values of each printing point to the wall thickness error potential energy values corresponding to the positions of the printing points to construct a wall thickness error potential energy field map; S4. Construct a target scanning density function according to the wall thickness error potential energy field map, solve the two-dimensional Monge-Ampère equation based on the target scanning density function to generate a path potential function, and generate a pseudo-potential streamline field according to the path potential function as the scanning path for the current printing layer; S5. After completing the printing of the current layer, jointly input the actual contour image, wall thickness data, pseudo-potential streamline field, and control parameters of the current layer into the wall thickness prediction neural network model to output the predicted wall thickness deviation values of each scanning unit in the next layer; S6. Update the wall thickness error map of the next layer according to the predicted wall thickness deviation values of each scanning unit in the next layer, and dynamically adjust the control parameters to complete the adaptive update of the control parameters; S7. Loop and execute steps S3 to S6 until the printing task is completed.

[0009] Optionally, the initial machining control parameter set includes the scanning path, laser power, scanning speed, scanning spacing, and spot size corresponding to each layer, and is generated by matching according to the cross-sectional geometric characteristics of each layer channel in the three-dimensional geometric model.

[0010] Optionally, the S2 specifically includes: S21. Under the condition of not applying the melting laser power, apply a variable-frequency laser scan with a frequency range of 10 Hz to 10 kHz to the combustion channel area to be printed, and control the laser power not to exceed 20% of the material melting threshold, so as to stimulate the thermal response change of the area to be printed; S22. Use an infrared thermal imager to collect the thermal response time series data of each scanning point during the laser scanning process, and record the temperature rise process and temperature change during the cooling process of each scanning point; S23. Process the thermal response time series data of each scanning point, extract the thermal response characteristic parameters, and the thermal response characteristic parameters include peak temperature, heating time, half-life, and temperature decay slope, to form a thermal response characteristic matrix; S24. Compare the thermal response characteristic matrix with a preset thermal behavior model point by point. The preset thermal behavior model uses the aluminum alloy powder response under the standard packing density as the reference curve library, and determines the thermal behavior deviation of each actual scanning point according to the principle of the minimum Euclidean distance; S25. Construct a two-dimensional difference image with the thermal behavior deviation value of each scanning point as the pixel gray value, and construct a thermal echo map in the form of a thermal map after normalizing the thermal behavior deviation. The thermal echo map reflects the change trend of the thermal absorption coefficient in the local area; S26. According to the change range of the thermal absorption coefficient of each scanning point in the thermal echo map, adjust the laser power, scanning speed, and spot diameter of the first layer in the corresponding area, generate the corrected machining control parameter set for the first layer and use it for the first layer printing.

[0011] Optionally, the S3 specifically includes: S31. Obtain the actual forming data after the completion of the previous layer printing. The actual forming data includes a three-dimensional contour image and actual wall thickness measurement data. The three-dimensional contour image is collected by an optical coherence tomography scanner, and the actual wall thickness measurement data is obtained through image reconstruction and calibration point spacing calculation; S32. Correspond the target design wall thickness of the current layer with the actual wall thickness measurement data at each printing point position one by one, calculate the wall thickness error value at each printing point, and form a wall thickness error distribution matrix; S33. Perform two-dimensional interpolation processing and Gaussian smoothing on the wall thickness error distribution matrix to obtain the wall thickness error map of the current layer. The wall thickness error map records the error size and spatial position of each printing point in the form of an image; S34. Set an error mapping function to map the wall thickness error values of each printing point to the wall thickness error potential energy values corresponding to the positions of the printing points. S35. Construct a wall thickness error potential energy field diagram based on the distribution of the wall thickness error potential energy values obtained from the error mapping function. The wall thickness error potential energy field diagram is represented in the form of a two-dimensional matrix. Each matrix element corresponds to the position of a printing point and records the wall thickness error potential energy value at the position of the printing point, forming a continuous spatial distribution data structure for describing the magnitude and directional change trend of the wall thickness deviation in each region.

[0012] Optionally, the error mapping function is specifically expressed as: ; Where represents the wall thickness error potential energy value at the printing point of the current layer, represents the wall thickness error value at the printing point of the current layer, , , , and represent preset regulation coefficients.

[0013] Optionally, the S4 specifically includes: S41. Normalize the wall thickness error potential energy field diagram to obtain a normalized wall thickness error potential energy value , with a value range between ; S42. Construct a target scanning density function based on the normalized wall thickness error potential energy value, such that is positively correlated with and satisfies the integral conservation constraint: ; Where represents the printing area of the current layer, represents the total amount of scanning path resources of the current layer; S43. Construct a Monge - Ampère equation with directional coupling and solve for the path potential function using the target scanning density function as the target term. The Monge - Ampère equation is: ; Where represents the determinant value, represents the Hessian matrix of the path potential function, represents the coupling coefficient, represents the directional coupling matrix, defined as: ; Among them, represents the wall thickness error gradient direction; S44. Set the boundary condition of the path potential function to the Neumann boundary condition, and specify that the normal derivative of the path potential function is zero on the boundary of the printing area to maintain the boundary continuity of the path field; S45. Solve the Monge-Ampère equation to obtain the two-dimensional distribution result of the path potential function ; S46. Calculate the weighted gradient field vector for scanning path guidance: ; Among them, represents the weighted gradient field vector, and represent the control weight coefficients, represents the gradient of the path potential function, represents the gradient of the normalized wall thickness error potential energy field; S47. Construct a pseudo-potential streamline field according to the direction distribution of the weighted gradient field vector as the scanning path of the current printing layer.

[0014] Optionally, the wall thickness prediction neural network model includes: A spatial feature encoding module for receiving and processing the actual contour image, wall thickness data, pseudo-potential streamline field and control parameters of the current layer, and extracting multi-scale spatial features through a two-dimensional convolutional encoder respectively; A time feature extraction module for loading the historical wall thickness error maps and control parameter sequences of multiple consecutive printing layers, and extracting the time evolution features of error changes by using a gated recurrent unit; A feature fusion module for splicing and fusing the multi-scale spatial features and time evolution features in the feature dimension, and enhancing the joint expression ability of local patterns and global trends through residual connections to generate fused joint features; A prediction generation module for generating a two-dimensional wall thickness deviation prediction matrix based on the fused joint features through a two-dimensional convolutional decoder, and the two-dimensional wall thickness deviation prediction matrix represents the wall thickness deviation prediction values of the next layer at each scanning unit.

[0015] Optionally, the specific content of S6 includes: Based on the wall thickness deviation prediction values of the next layer at each scanning unit, define a first threshold , a second threshold and a third threshold , and dynamically update each control parameter in the initial processing control parameter set of the next layer according to the set rules: If the predicted wall thickness deviation value is less than or equal to the first threshold , the predicted change rate of the wall thickness deviation value is less than or equal to the second threshold and the deviation fluctuation of two consecutive layers in the scanning unit is less than or equal to the third threshold , the control parameters remain unchanged and it is determined as a stable region; If any of the following conditions is met: the predicted wall thickness deviation value is greater than the first threshold , the predicted change rate of the wall thickness deviation value is greater than the second threshold , but the deviation fluctuation of two consecutive layers in the scanning unit is less than or equal to the third threshold , it is determined as a slightly corrected region, and the control parameters are adjusted based on the predicted wall thickness deviation value and the preset adjustment factor: ; wherein represents the adjusted laser power, represents the laser power before adjustment, represents the laser power adjustment factor, represents the predicted wall thickness deviation value; ; wherein represents the adjusted scanning speed, represents the scanning speed before adjustment, represents the scanning speed adjustment factor; ; wherein represents the adjusted scanning pitch, represents the scanning pitch before adjustment, represents the scanning pitch adjustment factor; ; wherein represents the adjusted spot size, represents the spot size before adjustment, represents the spot size adjustment factor; The scanning path remains unchanged in the slightly corrected region; If it simultaneously meets the conditions that the predicted wall thickness deviation value is greater than the first threshold , the predicted change rate of the wall thickness deviation value is greater than the second threshold and the deviation fluctuation of two consecutive layers in the scanning unit is greater than the third threshold , it is determined as a key corrected region, and the laser power, scanning speed, scanning pitch and spot size are adjusted by up to ±15% of the maximum correction amplitude, and a pseudo-potential streamline field is regenerated for the key corrected region to update the scanning path.

[0016] The beneficial effects of the present invention are as follows: (1) By introducing non-melting scanning of variable-frequency laser and constructing a thermal echo map before the first layer printing, it is possible to accurately identify the local thermophysical response differences of the powder bed before printing, effectively correct the control parameters of the first layer, and improve the consistency and stability of the printing initial conditions; (2) By adopting the wall thickness error potential field map constructed based on the actual forming data and solving the coupled directional Monge-Ampère equation, the linkage control of the scanning path and the local error spatial distribution is realized, so that the scanning trajectory dynamically tends to the low-potential error region, thereby suppressing the accumulation and diffusion of local wall thickness deviation; (3) By constructing a wall thickness prediction neural network model to synchronously extract wall thickness data, path field images and historical error evolution sequences, accurate prediction of wall thickness deviation across multiple levels and dimensions is realized, and the spatio-temporal perception ability of the prediction model is improved; (4) Based on the prediction results of wall thickness deviation, an error level region map is constructed to realize the point-by-point dynamic adaptive adjustment of control parameters such as laser power, scanning speed, scanning spacing and spot size, significantly enhancing the rapid response and closed-loop control ability of the manufacturing process to local errors; (5) A full-process dynamic regulation mechanism from thermal response modeling, error potential field analysis, path optimization, prediction feedback to control adjustment is constructed, effectively solving the problems of fixed path, static parameters and delayed feedback in the prior art, and realizing the systematic optimization of wall thickness control under high-complexity curved surface structures. Description of the Drawings

[0017] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, but do not constitute a limitation to the present invention. In the drawings: Figure 1 is the overall flowchart of a method for controlling the wall thickness uniformity of an additive manufacturing combustion channel in a space propulsion chamber proposed by the present invention; Figure 2 is the flowchart of generating the path potential function and the quasi-potential streamline field of the Monge-Ampère equation for a method for controlling the wall thickness uniformity of an additive manufacturing combustion channel in a space propulsion chamber proposed by the present invention; Figure 3 is the structural schematic diagram of the wall thickness prediction neural network model for a method for controlling the wall thickness uniformity of an additive manufacturing combustion channel in a space propulsion chamber proposed by the present invention. Detailed Embodiments

[0018] Now, the present invention will be further described in detail with reference to the drawings. These drawings are all simplified schematic diagrams, only illustrating the basic structure of the present invention in a schematic manner, so they only show the components related to the present invention.

[0019] ReferenceFigures 1-3 , a method for controlling the wall thickness uniformity of an additive manufacturing combustion channel in a space propulsion chamber, comprising the following steps: S1. Establish a three-dimensional geometric model of the combustion channel and perform layer slicing to obtain the structural information of each layer, and generate an initial processing control parameter set; S2. Before printing the first layer, use a variable-frequency laser to perform non-melting scanning on the area of the combustion channel to be printed and construct a thermal echo map, and correct the control parameters for the first layer in the initial processing control parameter set according to the thermal echo map; S3. Before printing each layer, generate a wall thickness error map for the current layer based on the actual forming data of the previous layer, map the wall thickness error values of each printing point to the wall thickness error potential energy values corresponding to the positions of the printing points, and construct a wall thickness error potential energy field map; S4. Construct a target scanning density function according to the wall thickness error potential energy field map, solve the two-dimensional Monge-Ampère equation based on the target scanning density function to generate a path potential function, and generate a quasi-potential streamline field according to the path potential function as the scanning path for the current printing layer; S5. After completing the printing of the current layer, jointly input the actual contour image, wall thickness data, quasi-potential streamline field and control parameters of the current layer into the wall thickness prediction neural network model, and output the predicted wall thickness deviation values of each scanning unit of the next layer; S6. Update the wall thickness error map of the next layer according to the predicted wall thickness deviation values of each scanning unit of the next layer, and dynamically adjust the control parameters to complete the adaptive update of the control parameters; S7. Loop through steps S3 to S6 until the printing task is completed.

[0020] The present invention establishes a closed-loop additive manufacturing control process for the combustion channel structure of a space propulsion chamber, combines pre-scanning thermal perception, construction of a wall thickness error potential energy field, path function optimization and linkage regulation of a depth prediction model, and forms a dynamic iterative mechanism from forming prediction to path planning and then to parameter adjustment. Through this method, the local error trend can be accurately identified before each printing layer, and a dynamic adaptive scanning path can be generated based on a mathematical physics model, so as to realize on-demand adjustment of the scanning density and energy input, thereby effectively suppressing the cross-layer accumulation of wall thickness deviation. This method not only significantly improves the uniformity and control granularity of the wall thickness of the curved surface, but also enhances the local sensitivity of the path response, and solves the problems of control hysteresis and parameter rigidity in complex wall structures.

[0021] In this embodiment, the initial processing control parameter set includes the scanning path, laser power, scanning speed, scanning spacing and spot size corresponding to each layer, and is generated by matching according to the cross-sectional geometric characteristics of each layer of the channel in the three-dimensional geometric model.

[0022] In this embodiment, the specific steps of S2 include: S21. Under the condition of not applying the melting laser power, a variable-frequency laser scanning with a frequency range of 10 Hz to 10 kHz is applied to the combustion channel area to be printed, and the laser power is controlled not to be higher than 20% of the material melting threshold, so as to stimulate the thermal response change of the area to be printed; S22. An infrared thermal imager is used to collect the thermal response time series data of each scanning point during the laser scanning process, and record the temperature rise process and temperature change during the cooling process of each scanning point; S23. Process the thermal response time series data of each scanning point, and extract the thermal response characteristic parameters. The thermal response characteristic parameters include peak temperature, heating time, half-life and temperature decay slope, and form a thermal response characteristic matrix; S24. Compare the thermal response characteristic matrix with a preset thermal behavior model point by point. The preset thermal behavior model uses the aluminum alloy powder response under the standard packing density as the reference curve library, and determines the thermal behavior deviation of each actual scanning point according to the principle of the minimum Euclidean distance; S25. Construct a two-dimensional difference image with the thermal behavior deviation value of each scanning point as the pixel gray value, and construct a thermal echo map through the form of the thermal map projection after normalizing the thermal behavior deviation. The thermal echo map reflects the change trend of the thermal absorption coefficient in the local area; S26. According to the change range of the thermal absorption coefficient of each scanning point in the thermal echo map, adjust the first-layer laser power, scanning speed and spot diameter of the corresponding area, generate a set of corrected first-layer processing control parameters and use them for the first-layer printing.

[0023] By introducing non-melting variable-frequency laser scanning and constructing a thermal echo map, a spatial tomography sensing mechanism for the local thermal physical state before printing is established, and high-resolution thermal response characteristics can be obtained without affecting the powder bed state. By using a standardized thermal behavior model to match the actual thermal response data point by point, a difference distribution map of the real-space thermal absorption characteristics can be obtained, and then the accurate correction of the first-layer processing parameters can be realized. This step significantly enhances the adaptability to the local powder thermal stability in the initial stage of printing, reduces the risk of energy input imbalance in the initial stage of forming, and helps to control the source of thermal distortion and the trend of error diffusion.

[0024] In this embodiment, the specific content of S3 includes: S31. Obtain the actual forming data after the completion of the previous layer of printing. The actual forming data includes a three-dimensional contour image and actual wall thickness measurement data. The three-dimensional contour image is collected by an optical coherence tomography scanner, and the actual wall thickness measurement data is obtained through image reconstruction and calibration point spacing calculation; S32. Correspond the target design wall thickness of the current layer with the actual wall thickness measurement data at each printing point position one by one, calculate the wall thickness error value at each printing point, and form a wall thickness error distribution matrix; S33. Perform two-dimensional interpolation and Gaussian smoothing on the wall thickness error distribution matrix to obtain the wall thickness error map of the current layer. The wall thickness error map records the error magnitudes and spatial positions of each printing point in the form of an image. S34. Set an error mapping function to map the wall thickness error values of each printing point to the wall thickness error potential energy values at the corresponding printing point positions. S35. According to the distribution of the wall thickness error potential energy values obtained from the error mapping function, construct a wall thickness error potential energy field map. The wall thickness error potential energy field map is represented in the form of a two-dimensional matrix, where each matrix element corresponds to a printing point position and records the wall thickness error potential energy value at the printing point position, forming a continuous spatial distribution data structure for describing the magnitudes and directional change trends of wall thickness deviations in each region.

[0025] The process of constructing the wall thickness error potential energy field provides a continuous spatial field basis for path optimization, and can transform discrete printing error data into a continuous control target with physical guiding significance. By mapping the wall thickness error to potential energy values and generating a spatial error field, a quantitative representation and directional description of local wall thickness abnormal regions are realized, effectively solving the problem that the traditional error control mode based on contour lines or profiles cannot reflect spatial gradient changes, and providing a more refined and differentiable input expression form for path function calculation and control parameter scheduling.

[0026] In this embodiment, the error mapping function is specifically expressed as: ; where represents the wall thickness error potential energy value at the printing point of the current layer, represents the wall thickness error value at the printing point of the current layer, , , , and represent preset regulation coefficients.

[0027] The adopted error mapping function introduces high-order regulation factors when setting, which can flexibly adjust the non-linear degree of the potential energy response to the error value, thereby endowing the error potential energy field with higher resolution ability and response dynamic range. By combining the exponential term, the inverse hyperbolic tangent term and the multiple slope control factor, the error mapping function has controllable characteristics in terms of boundary smoothing, central response enhancement and gradient transition, effectively suppressing the distortion risk caused by abnormal error points to the path guiding result, and enhancing the convergence of the path field and the practical processing feasibility.

[0028] In this embodiment, the specific content of S4 includes: S41. Normalize the wall thickness error potential energy field diagram to obtain the normalized wall thickness error potential energy value , with a value range within ; S42. Construct a target scan density function such that is positively correlated with and satisfies the integral conservation constraint: ; where represents the current layer printing area, represents the total amount of scan path resources in the current layer; S43. Construct a Monge - Ampère equation with directional coupling and solve for the path potential function using the target scan density function as the target term. The Monge - Ampère equation is: ; where represents the determinant value, represents the Hessian matrix of the path potential function, represents the coupling coefficient, represents the directional coupling matrix, defined as: ; where represents the wall thickness error gradient direction; S44. Set the boundary condition of the path potential function as the Neumann boundary condition, specifying that the normal derivative of the path potential function is zero on the boundary of the printing area to maintain the boundary continuity of the path field; S45. Solve the Monge - Ampère equation to obtain the two - dimensional distribution result of the path potential function ; S46. Calculate the weighted gradient field vector for scan path guidance: ; where represents the weighted gradient field vector, and represent the control weight coefficients, represents the gradient of the path potential function, represents the gradient of the normalized wall thickness error potential energy field; S47. Construct a pseudo - potential streamline field based on the direction distribution of the weighted gradient field vector as the scan path for the current printing layer.

[0029] Introduce the Monge - Ampère equation into the scanning path construction process. By setting the coupling term of the target scanning density function and the directional gradient, the generation of the potential function of the scanning path can be highly coupled with the error field distribution. The calculation of the path potential function not only satisfies the local density constraint but also guides the path trend to converge to the low - error region through boundary conditions and the weighted gradient field, realizing the adaptive regulation of the scanning trajectory in space. This path generation strategy has mathematical closure and physical rationality, significantly superior to traditional path raster methods or contour extension strategies, and shows stronger deformation adaptability and continuous stability in the control of the wall thickness inside the curved surface.

[0030] In this embodiment, the wall - thickness prediction neural network model includes: A spatial feature encoding module, which is used to receive and process the actual contour image, wall - thickness data, pseudo - potential streamline field, and control parameters of the current layer, and extract multi - scale spatial features through a two - dimensional convolutional encoder respectively; A temporal feature extraction module, which is used to load the historical wall - thickness error maps and control parameter sequences of multiple consecutive printing layers, and extract the temporal evolution features of the error changes by using a gated recurrent unit; A feature fusion module, which is used to splice and fuse the multi - scale spatial features and temporal evolution features in the feature dimension, and enhance the joint expression ability of local patterns and global trends through residual connections to generate the fused joint features; A prediction generation module, which is used to generate a two - dimensional wall - thickness deviation prediction matrix based on the fused joint features through a two - dimensional convolutional decoder. The two - dimensional wall - thickness deviation prediction matrix represents the predicted values of the wall - thickness deviation at each scanning unit of the next layer.

[0031] The wall - thickness prediction neural network model fully integrates the spatial structure features, control parameter information, and historical error evolution sequences in the printing process, and has a high - level context modeling ability. By extracting local geometric patterns through convolution, capturing temporal dynamic features through gated structures, and retaining residual connection paths in the fusion stage, the learning ability of the model for non - stationary and non - linear error change trends is effectively improved. The finally output two - dimensional prediction matrix corresponds one - to - one with the spatial scanning units, providing a high - resolution prediction basis for the precise dynamic adjustment of control parameters and realizing the efficient docking of the prediction model to the control logic.

[0032] In this embodiment, the specific content of S6 includes: Based on the predicted values of the wall - thickness deviation at each scanning unit of the next layer, define the first threshold , the second threshold and the third threshold , and dynamically update each control parameter in the initial processing control parameter set of the next layer according to the set rules: If the predicted value of the wall - thickness deviation satisfies being less than or equal to the first threshold The change rate of the predicted wall thickness deviation is less than or equal to the second threshold and the deviation fluctuation of two consecutive layers in the scanning unit is less than or equal to the third threshold , then the control parameters remain unchanged and it is determined as a stable area; If any one of the conditions is met: the predicted wall thickness deviation is greater than the first threshold , the change rate of the predicted wall thickness deviation is greater than the second threshold , but the deviation fluctuation of two consecutive layers in the scanning unit is less than or equal to the third threshold , then it is determined as a minor correction area, and the control parameters are adjusted based on the predicted wall thickness deviation and the preset adjustment factor: ; Among them, represents the adjusted laser power, represents the laser power before adjustment, represents the laser power adjustment factor, represents the predicted wall thickness deviation; ; Among them, represents the adjusted scanning speed, represents the scanning speed before adjustment, represents the scanning speed adjustment factor; ; Among them, represents the adjusted scanning pitch, represents the scanning pitch before adjustment, represents the scanning pitch adjustment factor; ; Among them, represents the adjusted spot size, represents the spot size before adjustment, represents the spot size adjustment factor; The scanning path remains unchanged in the minor correction area; If both the predicted wall thickness deviation is greater than the first threshold , the change rate of the predicted wall thickness deviation is greater than the second threshold and the deviation fluctuation of two consecutive layers in the scanning unit is greater than the third threshold , then it is determined as a key correction area, and the laser power, scanning speed, scanning pitch and spot size are adjusted by ±15% of the maximum correction amplitude limit, and a pseudo-potential streamline field is regenerated for the key correction area to update the scanning path.

[0033] By adopting a control parameter adjustment mechanism constructed based on the predicted wall thickness deviation, the point-by-point dynamic adaptive adjustment of laser power, scanning speed, scanning spacing, and spot size is realized. By setting multi-level response thresholds and change rate criteria, a three-level control state of a stable region, a slight correction region, and a key correction region is constructed, enabling the system to have a multi-level response ability. Further, in cooperation with the path reconstruction strategy, the scanning strategy of the error-sensitive area is updated, so as to effectively achieve the local compensation and global equilibrium adjustment of the error accumulation area, and significantly improve the accuracy and real-time performance of local quality control in complex surface printing.

[0034] Example 1: To verify the feasibility of the present invention in implementation, the present invention is applied to the laser additive manufacturing task of the hot end structure of a short-burn liquid oxygen-kerosene thrust chamber carried out by a certain aerospace engine research and development unit, and a typical vortex combustion channel section is selected as the test object. The wall thickness target of this section of the channel structure is 1.2 mm to 1.5 mm, and laser selective melting forming is carried out using aluminum alloy powder. It has typical variable-curvature inner walls and non-uniform energy flow density distribution characteristics, and is one of the hot end structures with the greatest current precision control difficulty.

[0035] In the traditional process plan, for this type of structure, the strategies of equal scanning spacing, constant laser power, and fixed speed are used. Although the configuration forming can be completed, local thickness deviations of more than ±80 μm often occur in the arc transition area and the wall thickness mutation area, seriously affecting the channel cooling capacity and combustion stability, and the rework rate is as high as more than 22%. By applying the wall thickness uniformity control method provided by the present invention, the overall process is intelligently upgraded.

[0036] First, in the printing preparation stage, by this method, a variable-frequency laser is introduced to perform non-melting scanning on the powder laying layer. The laser frequency is set to vary from 200 Hz to 2000 Hz, and the laser power is controlled at 15% of the original powder melting threshold. Combining the recording of the thermal response process by an infrared thermal imager, a complete thermal echo map is constructed. This map reveals the spatial distribution of the lag of thermal absorption response in local areas such as the diversion corner and the upper edge of the overhang, effectively guiding the differential correction of the first-layer parameters. The laser power in the area with uneven thermal response is reduced by 8%, and the scanning speed is increased by 5%, effectively improving the forming stability of the first layer.

[0037] Secondly, during the additive manufacturing process, before each layer is printed, the system automatically extracts the three-dimensional contour map and actual wall thickness data after the previous layer is formed, and constructs the wall thickness error map of the current layer. After Gaussian filtering and spatial interpolation processing of the error map, a wall thickness potential energy field map is generated. By setting the potential energy mapping function, the error range is mapped to a continuous field from 0 to 1. Based on this field, the directional Monge-Ampère path potential function is solved to generate a quasi-potential streamline field as the actual scanning path, forming a trajectory distribution that adaptively matches the error trend. During the printing process, the scanning trajectories in the high-potential error regions are significantly denser, and at the same time, the power is turned down, and the path tends to the regions with smaller errors, realizing local harmonic control.

[0038] After each layer of printing is completed, the system collects the actual control parameters, contour images, and generated path fields of this layer, and inputs the error sequences of multiple historical layers into the residual prediction network model to generate the wall thickness deviation prediction matrix of the next layer. The prediction accuracy reaches 93.7% on average in different regions, and the prediction deviation at local abnormal points is controlled within ±15μm, far superior to the traditional empirical model.

[0039] In the adaptive control parameter stage, according to the prediction results, a three-stage control threshold is set. The dynamic adjustment range of the scanning speed reaches ±12%, the control accuracy of the spot size reaches 20μm, and the adjustment step of the laser power is controlled at ±0.5W. The total printing process involves a total of 1185 layers of data collection and feedback, and the path adjustment rate reaches 83.5%. The control parameters are adjusted 3.6 times per layer on average.

[0040] In the final detection link, the wall thickness distribution data of the entire channel is obtained through a high-precision three-coordinate measuring system and a tomographic CT scanner. Statistics show that within the total forming length of 130mm, the proportion of the area with a wall thickness error within ±20μm reaches 92.4%, and the maximum error does not exceed ±38μm. Compared with the original process plan, the overall wall thickness control accuracy has been improved by 71.6%, and the rework rate has dropped to 4.3%. The total time used for the printing task is 27 hours and 48 minutes, and the consumable rate is controlled within an error range of ±3%.

[0041] In this embodiment, by applying the method of the present invention to the actual printing task of the complex combustion channel structure of the aerospace propulsion chamber, its effectiveness and engineering adaptability in high-precision wall thickness control are verified. Compared with the traditional printing strategy of fixed path and static parameters, this solution realizes a full-process closed-loop control mechanism of pre-printing thermal response prediction, in-printing path dynamic regulation, and post-printing error intelligent prediction. Experimental data show that this method significantly reduces the local wall thickness error, improves the spatial resolution of path control and the accuracy of parameter adjustment, successfully improves the wall thickness control accuracy to within ±20μm, and the rework rate drops by more than 80%, effectively solving the problems of path rigidity, error accumulation, and control lag in the existing technology.

[0042] In addition, this embodiment demonstrates the excellent adaptability of the method of the present invention in dealing with complex thermal load structures and regions with high-gradient wall thickness changes. By constructing a wall thickness error potential field and introducing a Monge-Ampère path optimization mechanism, the scanning path can intelligently tend to the region with smaller errors according to the real-time error distribution, effectively avoiding the regional aggregation of wall thickness deviations. At the same time, the wall thickness prediction neural network model that integrates the deep residual network and historical layer data not only improves the prediction accuracy but also provides a reliable basis for the feedforward adjustment of control parameters. The dynamic adjustment of control parameters at each layer ensures the optimal matching of energy distribution during the processing, significantly improving the delayed response problem of traditional printing methods in wall thickness control. This embodiment not only has replicability and generalizability but also provides practical technical support for the quality control of additive manufacturing of high-performance propulsion chambers, having important engineering application value and promotion significance.

[0043] As described above, only the preferred specific embodiments of the present invention are provided, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, making equivalent substitutions or changes, should be covered within the protection scope of the present invention.

Claims

1. A method for controlling the wall thickness uniformity of a combustion channel in an aerospace propulsion chamber by additive manufacturing, characterized in that, It includes the following steps: S1. Establish a three-dimensional geometric model of the combustion channel and perform layer-by-layer slicing to obtain the structural information of each layer, and generate an initial processing control parameter set; S2. Before printing the first layer, use a variable-frequency laser to perform non-melting scanning on the area of the combustion channel to be printed and construct a thermal echo map, and correct the control parameters for the first layer in the initial processing control parameter set according to the thermal echo map; S3. Before printing each layer, generate a wall thickness error map for the current layer based on the actual forming data of the previous layer, map the wall thickness error values of each printing point to the wall thickness error potential energy values corresponding to the positions of the printing points, and construct a wall thickness error potential energy field map; S4. Construct a target scanning density function according to the wall thickness error potential energy field map, solve the two-dimensional Monge-Ampère equation based on the target scanning density function to generate a path potential function, and generate a quasi-potential streamline field according to the path potential function as the scanning path for the current printing layer; S5. After completing the printing of the current layer, jointly input the actual contour image, wall thickness data, quasi-potential streamline field and control parameters of the current layer into the wall thickness prediction neural network model, and output the predicted wall thickness deviation values of each scanning unit of the next layer; S6. Update the wall thickness error map of the next layer according to the predicted wall thickness deviation values of each scanning unit of the next layer, and dynamically adjust the control parameters to complete the adaptive update of the control parameters; S7. Loop through steps S3 to S6 until the printing task is completed.

2. The method for controlling the wall thickness uniformity of a combustion channel in an aerospace propulsion chamber by additive manufacturing according to claim 1, characterized in that, The initial processing control parameter set includes the scanning path, laser power, scanning speed, scanning spacing and spot size corresponding to each layer, and is generated by matching according to the cross-sectional geometric characteristics of each layer of the channel in the three-dimensional geometric model.

3. A method for controlling the wall thickness uniformity of a combustion channel in an aerospace propulsion chamber by additive manufacturing according to claim 1, characterized in that, The specific content of S2 includes: S21. Under the condition of not applying the melting laser power, apply variable-frequency laser scanning with a frequency range of 10 Hz to 10 kHz to the area of the combustion channel to be printed, and control the laser power not to exceed 20% of the material melting threshold to stimulate the thermal response change of the area to be printed; S22. Use an infrared thermal imager to collect the thermal response time series data of each scanning point during the laser scanning process, and record the temperature rise process and temperature change during the cooling process of each scanning point; S23. Process the thermal response time series data of each scanning point, extract the thermal response characteristic parameters, and the thermal response characteristic parameters include peak temperature, heating time, half-life and temperature decay slope, to form a thermal response characteristic matrix; S24. Compare the thermal response characteristic matrix with a preset thermal behavior model point by point. The preset thermal behavior model uses the aluminum alloy powder response under the standard packing density as the reference curve library, and determines the thermal behavior deviation of each actual scanning point according to the principle of the smallest Euclidean distance; S25. Construct a two-dimensional difference image with the thermal behavior deviation value of each scanning point as the pixel gray value, and construct a thermal echo map through the thermal map projection form after normalizing the thermal behavior deviation. The thermal echo map reflects the change trend of the local area's thermal absorption coefficient; S26. Adjust the first-layer laser power, scanning speed, and spot diameter in the corresponding area according to the range of thermal absorption coefficient changes at each scanning point in the thermal echo map, generate a set of corrected processing control parameters for the first layer, and use them for the first-layer printing.

4. A method for controlling the wall thickness uniformity of a combustion channel in an aerospace propulsion chamber by additive manufacturing according to claim 1, characterized in that The specific steps of S3 are as follows: S31. Obtain the actual forming data after the completion of the previous layer of printing. The actual forming data includes a three-dimensional contour image and actual wall thickness measurement data. The three-dimensional contour image is collected by an optical coherence tomography scanner, and the actual wall thickness measurement data is obtained through image reconstruction and the calculation of the spacing between calibration points. S32. One-to-one correspond the target design wall thickness of the current layer with the actual wall thickness measurement data at each printing point position, calculate the wall thickness error value at each printing point, and form a wall thickness error distribution matrix. S33. Perform two-dimensional interpolation processing and Gaussian smoothing on the wall thickness error distribution matrix to obtain the wall thickness error map of the current layer. The wall thickness error map records the error magnitude and spatial position of each printing point in the form of an image. S34. Set an error mapping function to map the wall thickness error value of each printing point to the wall thickness error potential energy value at the corresponding printing point position. S35. According to the distribution of the wall thickness error potential energy values obtained by the error mapping function, construct a wall thickness error potential energy field map. The wall thickness error potential energy field map is represented in the form of a two-dimensional matrix, and each matrix element corresponds to a printing point position, recording the wall thickness error potential energy value at the printing point position, forming a continuous spatial distribution data structure for describing the magnitude and directional change trend of the wall thickness deviation in each area.

5. A method for controlling the wall thickness uniformity of a combustion channel in an aerospace propulsion chamber by additive manufacturing according to claim 4, characterized in that, The specific expression of the error mapping function is: ； Among them, represents the wall thickness error potential energy value at the current layer printing point ; represents the wall thickness error value at the current layer printing point ; , , , and represent preset regulation coefficients.

6. The method for controlling the wall thickness uniformity of an additive manufactured combustion channel in a space propulsion chamber according to claim 1, wherein The specific steps of S4 are as follows: S41. Normalize the wall thickness error potential energy field diagram to obtain the normalized wall thickness error potential energy value , with a value range within ; S42. Construct a target scanning density function based on the normalized wall thickness error potential value , such that is positively correlated with , and satisfies the integral conservation constraint: ； Among them, represents the printing area of the current layer, represents the total amount of scanning path resources of the current layer; S43. Construct the Monge - Ampère equation with directional coupling and solve the path potential function with the target scan density function as the target term , and the Monge - Ampère equation is: ； Among them, represents the determinant value, represents the Hessian matrix of the path potential function, represents the coupling coefficient, represents the directional coupling matrix, defined as: ； Among them, represents the direction of the wall thickness error gradient; S44. Set the boundary condition of the path potential function as the Neumann boundary condition, and specify that the normal derivative of the path potential function is zero on the boundary of the printing area to maintain the boundary continuity of the path field. S45. Solve the Monge - Ampère equation to obtain the path potential function and obtain the two - dimensional distribution result; S46. Calculate the weighted gradient field vector for scanning path guidance: ； Among them, represents the weighted gradient field vector, and represents the control weight coefficient, represents the gradient of the path potential function, represents the gradient of the normalized wall thickness error potential field; S47. Construct a quasi-potential streamline field according to the direction distribution of the weighted gradient field vector as the scanning path of the current printing layer.

7. A method for controlling the wall thickness uniformity of a combustion channel in an aerospace propulsion chamber by additive manufacturing according to claim 1, characterized in that, The wall thickness prediction neural network model includes: A spatial feature encoding module for receiving and processing the actual contour image, wall thickness data, quasi-potential streamline field, and control parameters of the current layer, and respectively extracting multi-scale spatial features through a two-dimensional convolutional encoder. A time feature extraction module for loading the historical wall thickness error maps and control parameter sequences of multiple consecutive printing layers, and using a gated recurrent unit to extract the time evolution features of the error changes. A feature fusion module for splicing and fusing the multi-scale spatial features and time evolution features in the feature dimension, and enhancing the joint expression ability of local patterns and global trends through residual connections to generate fused joint features. A prediction generation module for generating a two-dimensional wall thickness deviation prediction matrix based on the fused joint features through a two-dimensional convolutional decoder. The two-dimensional wall thickness deviation prediction matrix represents the wall thickness deviation prediction values at each scanning unit of the next layer.

8. A method for controlling the wall thickness uniformity of a combustion channel in an aerospace propulsion chamber by additive manufacturing according to claim 1, characterized in that, The specific steps of S6 are as follows: Define a first threshold value based on the predicted wall thickness deviation values of the next layer in each scanning unit , a second threshold value and a third threshold value , and dynamically update each control parameter in the initial processing control parameter set of the next layer according to the set rules: If the predicted value of the wall thickness deviation is less than or equal to the first threshold and the change rate of the predicted value of the wall thickness deviation is less than or equal to the second threshold and the deviation fluctuation of two consecutive layers in the scanning unit is less than or equal to the third threshold , the control parameters remain unchanged and it is determined as a stable region; If any of the following conditions is satisfied: the predicted value of wall thickness deviation is greater than the first threshold or the change rate of the predicted value of wall thickness deviation is greater than the second threshold , but the deviation fluctuation of the scanning unit is less than or equal to the third threshold for two consecutive layers , it is determined as a minor correction area, and the control parameters are adjusted based on the predicted value of wall thickness deviation and the preset adjustment factor, and the scanning path remains unchanged in the minor correction area; If the predicted value of wall thickness deviation is greater than the first threshold , the change rate of the predicted value of wall thickness deviation is greater than the second threshold and the deviation fluctuation of the scanning unit is greater than the third threshold for two consecutive layers , it is determined as a key correction area, and the laser power, scanning speed, scanning spacing, and spot size are adjusted by ±15% of the upper limit of the maximum correction amplitude, and the pseudo-potential streamline field of the key correction area is regenerated, and the scanning path is updated.

Images