A roughness optimization method for flow channel in additive manufacturing of aerospace propulsion chamber

Abstract

The present invention discloses a method for optimizing the roughness of a flow channel in additive manufacturing of a space propulsion chamber. The method comprises the following steps: setting a manufacturing parameter perturbation strategy, executing multiple sets of additive manufacturing experiments, obtaining a roughness formation sequence for the channel inner wall, generating a roughness variation trend using a gray prediction model, and combining roughness distribution data at multiple locations on the wall to guide a Dirichlet process to establish a spatial distribution model of the channel inner wall roughness. By fusing trend parameters with spatial features, real-time roughness data is collected during the manufacturing process, an error feedback mechanism is established, and linkage correction of the roughness prediction results is achieved. Finally, instructions for controlling the manufacturing path are generated to guide the machining process, thereby achieving precise control and optimization of the roughness of the flow channel inner wall in a space propulsion chamber. The present invention achieves linkage optimization by modeling the roughness evolution law, identifying the spatial distribution, and coordinating the manufacturing path.

Description

Technical Field

[0001] The present invention relates to the technical field of additive manufacturing process control, and in particular to a method for optimizing the roughness of a flow channel in additive manufacturing of an aerospace propulsion chamber. Background Art

[0002] With the widespread application of additive manufacturing technology in the aerospace field, the manufacturing accuracy and functional stability of the complex flow channel structure inside the aerospace propulsion chamber have become key issues of engineering concern. Especially in the propulsion channel structure working under high heat flux and high pressure coupling conditions, the surface roughness of the inner wall directly affects the fuel flow behavior and heat transfer performance, which in turn is related to the propulsion efficiency and thermal safety redundancy of the entire machine.

[0003] In existing technologies, the inner wall surfaces of flow channels formed by additive manufacturing processes such as metal selective laser melting often have uncontrollable microscopic roughness structures. Current process control methods mainly rely on empirical parameter settings and static error modeling, making it difficult to achieve dynamic understanding and feedforward prediction of the roughness formation mechanism. Specifically, existing technologies have obvious shortcomings in the following aspects:

[0004] 1. Lack of trend prediction capability: Current methods rely on a large amount of experimental data and fixed models to predict roughness changes, and lack a dynamic modeling mechanism suitable for small sample and highly coupled manufacturing environments.

[0005] 2. Model response rigidity: Traditional statistical regression or fitting methods have limited ability to characterize the spatial distribution of roughness and are unable to capture local abnormal behaviors during the multi-peak heterogeneous evolution of roughness.

[0006] 3. Feedback mechanism lag: The existing control system fails to establish an efficient prediction error feedback path, resulting in the manufacturing process being unable to dynamically correct and optimize the control path based on real-time roughness deviations.

[0007] 4. Insufficient parameter coupling processing: The formation of roughness is jointly affected by parameter disturbances such as laser power and scanning speed. The existing technology fails to establish a linkage mechanism between high-dimensional disturbances and roughness behavior.

[0008] Therefore, how to provide a method for optimizing the roughness of the flow channel in additive manufacturing of aerospace propulsion chambers is an urgent problem that needs to be solved by those skilled in the art. Summary of the Invention

[0009] One purpose of the present invention is to propose a roughness optimization method for the flow channel of the additive manufacturing of aerospace propulsion chamber. The present invention collects roughness data by manufacturing parameter perturbation experiment and constructs a gray prediction model based on the roughness of the flow channel. A roughness evolution trend model is constructed and the roughness spatial distribution model is generated in combination with the Dirichlet process. Real-time roughness monitoring and residual feedback mechanism are introduced for linkage optimization control, and finally the manufacturing path control instructions are generated. It has the advantages of high prediction accuracy, strong spatial adaptability, timely manufacturing feedback and fine roughness control.

[0010] A method for optimizing the roughness of a flow channel in additive manufacturing of an aerospace propulsion chamber according to an embodiment of the present invention includes the following steps:

[0011] Set the manufacturing parameter perturbation strategy, perform additive manufacturing experiments under multiple perturbation groups, and collect the corresponding channel inner wall roughness formation sequence to construct the roughness input-response relationship;

[0012] Constructing a grey prediction model based on the collected roughness formation sequence , a roughness time evolution model is generated in the form of differential equations to show the roughness changes with disturbance conditions. The roughness time evolution model is used to characterize the trend change behavior of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber caused by additive manufacturing parameter disturbances, and output roughness trend parameters and roughness prediction value sequences;

[0013] The roughness distribution data at different locations on the channel wall are collected and combined with the roughness trend parameter as a priori input to guide the Dirichlet process to establish a spatial probability distribution model of the roughness of the inner wall of the flow channel in the aerospace propulsion chamber.

[0014] Establish a fusion mechanism between roughness trend parameters and spatial probability distribution model, and output roughness trend parameters and roughness spatial probability distribution characteristics;

[0015] Collecting real-time roughness monitoring data during the additive manufacturing process, comparing the real-time roughness monitoring data with a roughness prediction value sequence, and calculating a prediction residual, wherein the prediction residual is used to measure the error between the roughness prediction and the actual formation;

[0016] A residual feedback mechanism is constructed based on the prediction residual, and the residual feedback mechanism is introduced into the fusion mechanism to drive the linkage optimization process;

[0017] Manufacturing path control instructions are generated based on the roughness trend parameters and roughness spatial probability distribution characteristics output by the fusion mechanism. The manufacturing path control instructions are used to dynamically adjust the scanning path density or energy input strategy of the flow channel area of ​​the aerospace propulsion chamber during the additive manufacturing process, so as to achieve the target optimization control of the roughness of the inner wall of the flow channel in the spatial position.

[0018] Optionally, the steps of setting a manufacturing parameter perturbation strategy, performing an additive manufacturing experiment, and collecting a channel inner wall roughness formation sequence include:

[0019] Selecting laser power, scanning speed, scanning spacing and forming layer thickness as manufacturing parameter perturbation dimensions, constructing a manufacturing parameter perturbation combination matrix, and the manufacturing parameter perturbation combination matrix serves as a component of the manufacturing parameter perturbation strategy;

[0020] Based on the manufacturing parameter perturbation combination matrix, additive forming experiments are sequentially performed on a metal laser selective melting additive manufacturing device, wherein the additive forming experiments generate samples including a flow channel structure of an aerospace propulsion chamber for different parameter combinations;

[0021] After completing each set of additive manufacturing experiments, surface topography measurement equipment was used to perform multi-position roughness measurements on the inner wall of the flow channel of the aerospace propulsion chamber of the sample. The roughness measurement area covered multiple characteristic locations of the inlet section, midsection, and outlet section.

[0022] The manufacturing parameter perturbation combination used in each set of additive forming experiments was paired with the corresponding measurement results of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber to construct a roughness formation sequence, which served as the input data basis for the grey prediction model GM(1,1) modeling process.

[0023] Optionally, the grey prediction model is constructed based on the collected roughness formation sequence , a roughness time evolution model is generated in the form of a differential equation, in which the roughness changes with the disturbance conditions, and the roughness trend parameters and roughness prediction value sequence are output. Specifically, the following are included:

[0024] The roughness formation sequence is constructed as the original time series data of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber. Based on the preset time window division rule, a segmented adjustable accumulation generation operation is performed to form multiple roughness gray generation subsequences.

[0025] Based on each roughness grey generation subsequence, the corresponding grey prediction submodel is established respectively , and in each grey prediction sub-model Generate a set of roughness trend characteristic parameters for trend analysis;

[0026] In each grey prediction sub-model In , a non-uniformly weighted background value sequence is constructed based on the roughness gray generator sequence:

[0027] ;

[0028] in, For the Item non-uniformly weighted background value sequence, Generate subsequences for roughness gray, is the weight coefficient, and its value range is , is the sequence index;

[0029] Construct each grey prediction sub-model based on the non-uniformly weighted background value sequence The differential prediction equation is:

[0030] ;

[0031] in, Generate subsequences for roughness gray, and is the estimated parameter of the grey prediction sub-model GM(1,1), and the parameter identification is performed by the least square method;

[0032] Each grey prediction sub-model The output roughness prediction value sequence is summarized and used as the roughness prediction input data for the residual feedback mechanism;

[0033] Each grey prediction sub-model The extracted roughness trend characteristic parameters are jointly integrated to form the roughness trend parameter, which is used as the trend priori input variable in the Dirichlet process roughness modeling module.

[0034] Optionally, the step of collecting roughness distribution data at different locations on the inner wall of the flow channel of the aerospace propulsion chamber, combining the roughness trend parameter as a priori input, and guiding the Dirichlet process to establish a spatial probability distribution model of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber includes:

[0035] The inner wall of the flow channel in the aerospace propulsion chamber is divided into multiple spatial measurement areas. Three types of roughness indicators, namely arithmetic mean roughness, maximum profile height, and root mean square roughness, are collected in each spatial measurement area. The three-dimensional spatial position coordinates corresponding to each set of roughness indicators are recorded to construct the roughness distribution data of the inner wall of the flow channel in the aerospace propulsion chamber.

[0036] The roughness distribution data of the inner wall of the flow channel of the aerospace propulsion chamber is used as the input of the non-parametric Bayesian model. The Dirichlet process adopts the multi-dimensional Gaussian distribution as the base distribution, and the concentration parameter is set to control the non-parametric prior strength of the number of clusters.

[0037] In the Dirichlet process modeling, the initial state is an empty cluster. During the sampling process, each new data point takes the current three-dimensional spatial position coordinates and roughness index as input, and calculates the likelihood value of the distribution parameters of each existing cluster. The distribution parameters of the cluster include the mean vector and covariance matrix of the roughness index.

[0038] According to the distribution parameters and likelihood value of each cluster and the corresponding cluster size, the concentration parameter is combined to calculate the belonging probability, perform cluster belonging sampling, and determine whether to generate a new roughness sub-distribution cluster;

[0039] In each sampling round, the roughness trend parameter output by the grey prediction model GM(1,1) is introduced as a priori adjustment variable to update the cluster distribution weight so that the roughness evolution trend and spatial distribution structure are adjusted synchronously.

[0040] After sampling convergence, all roughness sub-distribution clusters are merged to output the spatial probability distribution model of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber, which is used to describe the multi-peak distribution state and trend-guiding behavior of the roughness in different spatial regions.

[0041] Optionally, the fusion mechanism for establishing the roughness trend parameter and the spatial probability distribution model specifically includes establishing a mapping relationship between the roughness trend parameter and each roughness sub-distribution cluster in the spatial probability distribution model, calculating the residual between the roughness trend parameter and the mean of the corresponding roughness sub-distribution cluster in each mapping pair, and constructing a fusion weight adjustment function based on the residual, dynamically correcting the probability weights of each roughness sub-distribution cluster in the spatial probability distribution model of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber, and iteratively updating the roughness trend parameter using the residual as a feedback variable, thereby realizing a bidirectional dynamic coupling fusion between the grey prediction model GM(1,1) and the spatial probability distribution model of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber.

[0042] Optionally, collecting real-time roughness monitoring data during the additive manufacturing process, comparing the real-time roughness monitoring data with a roughness prediction value sequence, and calculating the prediction residual specifically include:

[0043] During the additive manufacturing process of the flow channel in aerospace propulsion chambers, a confocal laser measurement device was used to measure the inner wall of the flow channel online based on a preset scanning path and time interval. Three types of roughness indicators, including arithmetic mean roughness, maximum profile height, and root mean square roughness, were collected, along with the sampling time index and three-dimensional spatial position coordinates, forming real-time roughness monitoring data for the inner wall of the flow channel in aerospace propulsion chambers.

[0044] The real-time roughness monitoring data of the inner wall of the flow channel of the aerospace propulsion chamber is reorganized into a time series structure according to the time index, and a one-to-one correspondence is made with the roughness prediction value sequence of the inner wall of the flow channel of the aerospace propulsion chamber in chronological order. The three types of predicted roughness indicators are matched with the three types of measured roughness indicators at each sampling moment.

[0045] At each sampling time point, the numerical differences between the predicted values ​​and the measured values ​​of the three types of roughness, namely, arithmetic mean roughness, maximum profile height and root mean square roughness, are calculated as the prediction residuals at the corresponding time point.

[0046] The prediction residuals calculated at all sampling time points are constructed as a roughness prediction residual data set of the inner wall of the flow channel of the aerospace propulsion chamber. The roughness prediction residual data set is used as the input of the fusion mechanism to drive the dynamic correction and update of the parameters of the grey prediction model GM(1,1) and the spatial probability distribution model of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber.

[0047] Optionally, constructing a residual feedback mechanism based on the prediction residual and introducing the residual feedback mechanism into the fusion mechanism to drive the linkage optimization process specifically includes:

[0048] Based on the real-time roughness monitoring data and the roughness prediction value series of the inner wall of the flow channel of the aerospace propulsion chamber, the roughness prediction values ​​and the real-time roughness monitoring values ​​are paired point by point according to the time index. The arithmetic mean roughness residual, maximum profile height residual and root mean square roughness residual at each time point are calculated, and a residual time series is generated.

[0049] Segment the residual time series using a sliding window, calculate the mean, standard deviation, and trend slope of the residual series in each window, and construct a residual feedback vector;

[0050] The residual feedback vector is used as the input of the residual feedback mechanism, acting on the roughness trend parameter generation process and the iterative process of the roughness spatial probability distribution model of the inner wall of the flow channel of the aerospace propulsion chamber, forming a dual-channel regulation mechanism.

[0051] In the process of generating roughness trend parameters, the residual feedback mechanism is used to adjust the weight distribution structure of the non-uniformly weighted background value construction method in the grey prediction model GM(1,1) and the iterative update step size of the model parameters;

[0052] In the spatial probability distribution model of the inner wall roughness of the flow channel in aerospace propulsion chambers, the residual feedback mechanism is used to adjust the sampling and new cluster generation thresholds of each roughness sub-distribution cluster, and dynamically correct the cluster distribution weights of the spatial distribution.

[0053] Through the dual-path injection method of the residual feedback mechanism, a linkage optimization process is constructed, so that the roughness trend parameters and the spatial probability distribution characteristics of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber are collaboratively updated according to the roughness prediction residual.

[0054] Optionally, the step of generating manufacturing path control instructions based on the roughness trend parameters and spatial probability distribution characteristics output by the fusion mechanism includes:

[0055] The roughness trend parameters are expanded by time index, and the roughness change rate, fluctuation amplitude and local evolution trend corresponding to different manufacturing time periods are extracted to form the time dynamic control factor in the manufacturing path control instruction;

[0056] The spatial probability distribution characteristics of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber are regionally decomposed, and the spatial position center, covariance matrix and roughness distribution weight of each roughness sub-distribution cluster are extracted to form the spatial distribution control factor in the manufacturing path control instruction.

[0057] Based on the time dynamic control factor and the spatial distribution control factor, a manufacturing path control instruction parameter set is constructed. The manufacturing path control instruction includes:

[0058] Scan path structure parameters, including the three-dimensional coordinate sequence of the scanning segment, the scanning path direction vector and the scanning segment jump strategy;

[0059] Laser power control parameters, including roughness trend control coefficient and laser power base value and adjustment range under spatial distribution mapping;

[0060] Scanning speed control parameters, including a scanning speed function based on the roughness trend change rate, defining the scanning start speed, maximum speed, and speed gradient change form;

[0061] Scan spacing control parameters, including scan overlap, interlayer offset, and path spacing adjustment corresponding to changes in roughness spatial density;

[0062] The parameters in the manufacturing path control instruction parameter set are jointly adjusted according to the intersection area of ​​the roughness trend parameter and the spatial probability distribution characteristics of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber. A fusion control mapping model is constructed to generate manufacturing path control instruction sets for different manufacturing areas and time periods.

[0063] The manufacturing path control instruction set is converted into a manufacturing control file format that can be recognized by the aerospace propulsion chamber flow channel additive manufacturing equipment, and transmitted to the control system to complete the manufacturing execution.

[0064] The beneficial effects of the present invention are:

[0065] (1) The present invention constructs a sequence of manufacturing parameter disturbance combinations and roughness formation, uses the grey prediction model GM(1,1) to establish a roughness time evolution trend model, and introduces a piecewise adjustable accumulation and non-uniformly weighted background value construction mechanism to achieve interpretable modeling of the roughness evolution law and significantly improve the prediction ability of the roughness change trend.

[0066] (2) The present invention collects multi-region and multi-dimensional roughness index data of the inner wall of the flow channel of the aerospace propulsion chamber, guides the Dirichlet process to establish a spatial probability distribution model, and introduces roughness trend parameters for trend prior control, thereby realizing multi-peak modeling and trend guidance of roughness spatial distribution, and significantly enhancing the adaptability of the model to complex spatial roughness states.

[0067] (3) The present invention dynamically combines the roughness time trend parameter with the spatial probability distribution model by designing a fusion mechanism. On this basis, the real-time roughness monitoring data is introduced to compare with the roughness prediction value sequence to calculate the residual. The residual feedback mechanism is constructed and introduced into the trend prediction and spatial modeling process, realizing a two-way feedback control mechanism for the manufacturing process and improving the real-time response capability and model adaptability of the system.

[0068] (4) The present invention finally uses the fused roughness evolution trend and spatial distribution characteristics as input, and generates manufacturing path control instructions in combination with the aerospace propulsion chamber structure manufacturing requirements. The instruction structure includes spatial position index, manufacturing parameter fine-tuning items and process adjustment weights, which realizes the fine control of the additive manufacturing process and effectively improves the roughness control accuracy and consistency of the final part. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] The accompanying 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 and do not constitute a limitation of the present invention. In the accompanying drawings:

[0070] Figure 1 This is a flow chart of a roughness optimization method for additively manufactured flow channels in aerospace propulsion chambers proposed by the present invention. DETAILED DESCRIPTION

[0071] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0072] refer to Figure 1 A method for optimizing the roughness of a flow channel in additive manufacturing of an aerospace propulsion chamber comprises the following steps:

[0073] Set the manufacturing parameter perturbation strategy, perform additive manufacturing experiments under multiple perturbation groups, and collect the corresponding channel inner wall roughness formation sequence to construct the roughness input-response relationship;

[0074] Constructing a grey prediction model based on the collected roughness formation sequence , a roughness time evolution model is generated in the form of differential equations to show the roughness changes with disturbance conditions. The roughness time evolution model is used to characterize the trend change behavior of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber caused by additive manufacturing parameter disturbances, and output roughness trend parameters and roughness prediction value sequences;

[0075] The roughness distribution data at different locations on the channel wall are collected and combined with the roughness trend parameter as a priori input to guide the Dirichlet process to establish a spatial probability distribution model of the roughness of the inner wall of the flow channel in the aerospace propulsion chamber.

[0076] Establish a fusion mechanism between roughness trend parameters and spatial probability distribution model, and output roughness trend parameters and roughness spatial probability distribution characteristics;

[0077] Collecting real-time roughness monitoring data during the additive manufacturing process, comparing the real-time roughness monitoring data with a roughness prediction value sequence, and calculating a prediction residual, wherein the prediction residual is used to measure the error between the roughness prediction and the actual formation;

[0078] A residual feedback mechanism is constructed based on the prediction residual, and the residual feedback mechanism is introduced into the fusion mechanism to drive the linkage optimization process;

[0079] Manufacturing path control instructions are generated based on the roughness trend parameters and roughness spatial probability distribution characteristics output by the fusion mechanism. The manufacturing path control instructions are used to dynamically adjust the scanning path density or energy input strategy of the flow channel area of ​​the aerospace propulsion chamber during the additive manufacturing process, so as to achieve the target optimization control of the roughness of the inner wall of the flow channel in the spatial position.

[0080] In this embodiment, the steps of setting a manufacturing parameter perturbation strategy, performing an additive manufacturing experiment, and collecting channel inner wall roughness formation sequence include:

[0081] Selecting laser power, scanning speed, scanning spacing and forming layer thickness as manufacturing parameter perturbation dimensions, constructing a manufacturing parameter perturbation combination matrix, and the manufacturing parameter perturbation combination matrix serves as a component of the manufacturing parameter perturbation strategy;

[0082] Based on the manufacturing parameter perturbation combination matrix, additive forming experiments are sequentially performed on a metal laser selective melting additive manufacturing device, wherein the additive forming experiments generate samples including a flow channel structure of an aerospace propulsion chamber for different parameter combinations;

[0083] After completing each set of additive manufacturing experiments, surface topography measurement equipment was used to perform multi-position roughness measurements on the inner wall of the flow channel of the aerospace propulsion chamber of the sample. The roughness measurement area covered multiple characteristic locations of the inlet section, midsection, and outlet section.

[0084] The manufacturing parameter perturbation combination used in each set of additive forming experiments was paired with the corresponding measurement results of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber to construct a roughness formation sequence, which served as the input data basis for the grey prediction model GM(1,1) modeling process.

[0085] In this embodiment, a manufacturing parameter disturbance combination matrix is ​​constructed by selecting laser power, scanning speed, scanning spacing and forming layer thickness as disturbance dimensions, and forming experiments are carried out group by group on a metal laser selective melting additive manufacturing device to prepare a sample with a flow channel structure of an aerospace propulsion chamber. The roughness of the inner wall of the sample is measured at multiple characteristic positions such as the air inlet section, middle section and outlet section using a surface morphology measurement device, and multidimensional indicators such as the arithmetic mean roughness are extracted. The correspondence between the manufacturing parameters and the roughness data is constructed, and finally a roughness formation sequence is formed as the basic data for modeling a grey prediction model, effectively establishing a mapping structure between the disturbance parameters and the roughness response.

[0086] In this embodiment, the grey prediction model is constructed based on the roughness formation sequence collected. , a roughness time evolution model is generated in the form of a differential equation, in which the roughness changes with the disturbance conditions, and the roughness trend parameters and roughness prediction value sequence are output. Specifically, the following are included:

[0087] The roughness formation sequence is constructed as the original time series data of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber. Based on the preset time window division rule, a segmented adjustable accumulation generation operation is performed to form multiple roughness gray generation subsequences.

[0088] Based on each roughness grey generation subsequence, the corresponding grey prediction submodel is established respectively , and in each grey prediction sub-model Generate a set of roughness trend characteristic parameters for trend analysis;

[0089] In each grey prediction sub-model In , a non-uniformly weighted background value sequence is constructed based on the roughness gray generator sequence:

[0090] ;

[0091] in, For the Item non-uniformly weighted background value sequence, Generate subsequences for roughness gray, is the weight coefficient, and its value range is , is the sequence index;

[0092] This formula is used to construct a non-uniformly weighted background value sequence in the grey prediction model. Its core idea is to form a background value sequence that is more in line with the actual change trend by taking a weighted average of adjacent data points in the roughness grey generation subsequence. Unlike the traditional weighted processing, this formula introduces a weight coefficient to perform a non-balanced combination of the values ​​at the current moment and the previous moment, so that the background value can reflect the dynamic change characteristics of the time series during model training. This method improves the model's responsiveness to trend changes and reduces the model's dependence on data stationarity by enhancing the sensitivity to changes within the sequence. It has strong versatility and adaptability. In terms of creativity, this method adjusts the modeling sensitivity by introducing a non-uniformly weighted mechanism, achieving a flexible expression of the roughness evolution law in the time series modeling process, overcoming the problem of reduced prediction accuracy of traditional grey modeling under complex disturbance conditions, and having a significant technological breakthrough.

[0093] Construct each grey prediction sub-model based on the non-uniformly weighted background value sequence The differential prediction equation is:

[0094] ;

[0095] in, Generate subsequences for roughness gray, and is the estimated parameter of the grey prediction sub-model GM(1,1), and the parameter identification is performed by the least square method;

[0096] This formula represents the differential prediction form of the grey prediction sub-model constructed based on the non-uniformly weighted background value sequence. Its principle is to describe the dynamic process of the roughness generation sub-sequence changing over time by constructing a first-order linear differential equation. The two parameters to be estimated in the model represent the attenuation factor and external driving term of the system evolution. The least squares method is used to accurately fit the parameters, so that the model can capture the changing trend of the data and the disturbance response behavior. This method not only continues the grey system's ability to model a small number of samples, but also introduces a differential prediction structure on this basis, realizing continuous modeling and fine-grained characterization of the roughness change process, and improving the prediction resolution and dynamic adaptability of the model under disturbance drive. In terms of creativity, by introducing non-uniformly weighted background values ​​into grey differential modeling, the transformation from static weighted estimation to dynamic evolution modeling is realized, and the generalization ability and prediction stability of the model in complex manufacturing scenarios are enhanced.

[0097] Each grey prediction sub-model The output roughness prediction value sequence is summarized and used as the roughness prediction input data for the residual feedback mechanism;

[0098] Each grey prediction sub-model The extracted roughness trend characteristic parameters are jointly integrated to form the roughness trend parameter, which is used as the trend priori input variable in the Dirichlet process roughness modeling module.

[0099] This embodiment constructs the roughness formation sequence of the inner wall of the flow channel of the aerospace propulsion chamber as time series data, and performs a segmented adjustable accumulation operation within a preset time window to generate multiple roughness gray generation subsequences, and establishes gray prediction sub-models GM(1,1) respectively. In each model, a non-uniformly weighted background value construction method is introduced and a differential prediction equation is established to achieve detailed modeling and dynamic tracking of the roughness evolution trend. The roughness prediction value output by each sub-model is used in the subsequent residual feedback mechanism, and the trend characteristic parameters are integrated to form the roughness trend parameters and input into the Dirichlet process modeling process. This method improves the prediction model's ability to characterize the dynamics of roughness evolution under changing disturbance conditions through multi-segment modeling, and enhances the system's trend recognition accuracy for uncertain manufacturing disturbances.

[0100] In this embodiment, the steps of collecting roughness distribution data at different locations on the inner wall of the flow channel of the aerospace propulsion chamber, combining the roughness trend parameter as a priori input, and guiding the Dirichlet process to establish a spatial probability distribution model of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber include:

[0101] The inner wall of the flow channel in the aerospace propulsion chamber is divided into multiple spatial measurement areas. Three types of roughness indicators, namely arithmetic mean roughness, maximum profile height, and root mean square roughness, are collected in each spatial measurement area. The three-dimensional spatial position coordinates corresponding to each set of roughness indicators are recorded to construct the roughness distribution data of the inner wall of the flow channel in the aerospace propulsion chamber.

[0102] The roughness distribution data of the inner wall of the flow channel of the aerospace propulsion chamber is used as the input of the non-parametric Bayesian model. The Dirichlet process adopts the multi-dimensional Gaussian distribution as the base distribution, and the concentration parameter is set to control the non-parametric prior strength of the number of clusters.

[0103] In the Dirichlet process modeling, the initial state is an empty cluster. During the sampling process, each new data point takes the current three-dimensional spatial position coordinates and roughness index as input, and calculates the likelihood value of the distribution parameters of each existing cluster. The distribution parameters of the cluster include the mean vector and covariance matrix of the roughness index.

[0104] According to the distribution parameters and likelihood value of each cluster and the corresponding cluster size, the concentration parameter is combined to calculate the belonging probability, perform cluster belonging sampling, and determine whether to generate a new roughness sub-distribution cluster;

[0105] In each sampling round, the roughness trend parameter output by the grey prediction model GM(1,1) is introduced as a priori adjustment variable to update the cluster distribution weight so that the roughness evolution trend and spatial distribution structure are adjusted synchronously.

[0106] After sampling convergence, all roughness sub-distribution clusters are merged to output the spatial probability distribution model of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber, which is used to describe the multi-peak distribution state and trend-guiding behavior of the roughness in different spatial regions.

[0107] This implementation divides the inner wall of the flow channel in the aerospace propulsion chamber into multiple spatial measurement areas, collects multidimensional indicators including arithmetic mean roughness, maximum profile height, and root mean square roughness, and constructs roughness distribution data based on three-dimensional spatial coordinates. Using the Dirichlet process, a multidimensional Gaussian distribution and concentration parameters are introduced for non-parametric Bayesian modeling to establish an adaptive clustering model for roughness. During continuous sampling, cluster affiliation is dynamically adjusted based on the likelihood relationship between the current data point and the existing distribution cluster and the prior control factor. In each iteration, trend parameters generated by the gray prediction model GM(1,1) are introduced to synchronously adjust cluster weights. Ultimately, a roughness spatial probability distribution model is formed that integrates trend and spatial structure. This model accurately depicts the multimodal distribution and dynamic evolution of the flow channel inner wall roughness, improving the model's ability to identify local roughness anomalies in complex manufacturing environments.

[0108] In this embodiment, the fusion mechanism for establishing the roughness trend parameter and the spatial probability distribution model specifically includes establishing a mapping relationship between the roughness trend parameter and each roughness sub-distribution cluster in the spatial probability distribution model, calculating the residual between the roughness trend parameter and the mean of the corresponding roughness sub-distribution cluster in each mapping pair, and constructing a fusion weight adjustment function based on the residual, dynamically correcting the probability weight of each roughness sub-distribution cluster in the spatial probability distribution model of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber, and iteratively updating the roughness trend parameter using the residual as a feedback variable, thereby realizing a bidirectional dynamic coupling fusion between the grey prediction model GM(1,1) and the spatial probability distribution model of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber.

[0109] This embodiment establishes a one-to-one mapping relationship between the roughness trend parameter and each roughness sub-distribution cluster in the roughness spatial probability distribution model of the inner wall of the flow channel of the aerospace propulsion chamber, calculates the residual between the trend parameter and the sub-cluster mean in each mapping pair, and constructs a fusion weight adjustment function to drive the dynamic correction of the sub-distribution cluster probability weight with the residual. At the same time, the residual feedback is used for the iterative update of the trend parameter, realizing a two-way control coupling between the grey prediction model and the roughness space modeling, thereby establishing a dynamic consistency relationship between the roughness evolution trend and the spatial characteristics, thereby improving the accuracy of roughness modeling and the response sensitivity of manufacturing control.

[0110] In this embodiment, the real-time roughness monitoring data is collected during the additive manufacturing process, the real-time roughness monitoring data is compared with the roughness prediction value sequence, and the prediction residual is calculated specifically including:

[0111] During the additive manufacturing process of the flow channel in aerospace propulsion chambers, a confocal laser measurement device was used to measure the inner wall of the flow channel online based on a preset scanning path and time interval. Three types of roughness indicators, including arithmetic mean roughness, maximum profile height, and root mean square roughness, were collected, along with the sampling time index and three-dimensional spatial position coordinates, forming real-time roughness monitoring data for the inner wall of the flow channel in aerospace propulsion chambers.

[0112] The real-time roughness monitoring data of the inner wall of the flow channel of the aerospace propulsion chamber is reorganized into a time series structure according to the time index, and a one-to-one correspondence is made with the roughness prediction value sequence of the inner wall of the flow channel of the aerospace propulsion chamber in chronological order. The three types of predicted roughness indicators are matched with the three types of measured roughness indicators at each sampling moment.

[0113] At each sampling time point, the numerical differences between the predicted values ​​and the measured values ​​of the three types of roughness, namely, arithmetic mean roughness, maximum profile height and root mean square roughness, are calculated as the prediction residuals at the corresponding time point.

[0114] The prediction residuals calculated at all sampling time points are constructed as a roughness prediction residual data set of the inner wall of the flow channel of the aerospace propulsion chamber. The roughness prediction residual data set is used as the input of the fusion mechanism to drive the dynamic correction and update of the parameters of the grey prediction model GM(1,1) and the spatial probability distribution model of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber.

[0115] This embodiment uses a confocal laser measurement device to collect the roughness data of the inner wall of the channel in real time according to a preset path and time interval during the additive manufacturing process of the flow channel in the aerospace propulsion chamber, and extracts three types of indicators: arithmetic mean roughness, maximum profile height and root mean square roughness. After constructing a time series structure, the residuals are matched point by point with the predicted values ​​to form a roughness prediction residual data set, which is further used as feedback input to drive the dynamic correction of the gray prediction model and the roughness space probability distribution model, thereby realizing real-time perception and model adjustment of roughness changes in the manufacturing process, significantly enhancing the system's responsiveness to manufacturing disturbances and the prediction correction accuracy, and effectively improving the intelligence level of additive manufacturing roughness control and the consistency of parts.

[0116] In this embodiment, the residual feedback mechanism is constructed based on the prediction residual, and the residual feedback mechanism is introduced into the fusion mechanism to drive the linkage optimization process, specifically including:

[0117] Based on the real-time roughness monitoring data and the roughness prediction value series of the inner wall of the flow channel of the aerospace propulsion chamber, the roughness prediction values ​​and the real-time roughness monitoring values ​​are paired point by point according to the time index. The arithmetic mean roughness residual, maximum profile height residual and root mean square roughness residual at each time point are calculated, and a residual time series is generated.

[0118] Segment the residual time series using a sliding window, calculate the mean, standard deviation, and trend slope of the residual series in each window, and construct a residual feedback vector;

[0119] The residual feedback vector is used as the input of the residual feedback mechanism, acting on the roughness trend parameter generation process and the iterative process of the roughness spatial probability distribution model of the inner wall of the flow channel of the aerospace propulsion chamber, forming a dual-channel regulation mechanism.

[0120] In the process of generating roughness trend parameters, the residual feedback mechanism is used to adjust the weight distribution structure of the non-uniformly weighted background value construction method in the grey prediction model GM(1,1) and the iterative update step size of the model parameters;

[0121] In the spatial probability distribution model of the inner wall roughness of the flow channel in aerospace propulsion chambers, the residual feedback mechanism is used to adjust the sampling and new cluster generation thresholds of each roughness sub-distribution cluster, and dynamically correct the cluster distribution weights of the spatial distribution.

[0122] Through the dual-path injection method of the residual feedback mechanism, a linkage optimization process is constructed, so that the roughness trend parameters and the spatial probability distribution characteristics of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber are collaboratively updated according to the roughness prediction residual.

[0123] This implementation method generates multi-class residual time series by pairing the real-time roughness monitoring data of the inner wall of the flow channel of the aerospace propulsion chamber with the roughness prediction value sequence point by point, and uses a sliding window to perform statistical analysis on the residual sequence to construct a residual feedback vector, which is further injected into the grey prediction model GM(1,1) and the roughness spatial probability distribution model as feedback input, respectively. This realizes the dynamic adjustment of the non-uniformly weighted background value construction strategy and the model parameter update step, as well as the adaptive regulation of the distribution cluster sampling mechanism and the generation threshold, thereby constructing a trend-space dual-path linkage optimization process for the disturbance environment of the manufacturing process, realizing the synchronous update of the roughness evolution trend and the spatial distribution structure, and significantly improving the robustness and control accuracy of the model in an uncertain manufacturing environment.

[0124] In this embodiment, the step of generating manufacturing path control instructions based on the roughness trend parameters and spatial probability distribution characteristics output by the fusion mechanism includes:

[0125] The roughness trend parameters are expanded by time index, and the roughness change rate, fluctuation amplitude and local evolution trend corresponding to different manufacturing time periods are extracted to form the time dynamic control factor in the manufacturing path control instruction;

[0126] The spatial probability distribution characteristics of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber are regionally decomposed, and the spatial position center, covariance matrix and roughness distribution weight of each roughness sub-distribution cluster are extracted to form the spatial distribution control factor in the manufacturing path control instruction.

[0127] Based on the time dynamic control factor and the spatial distribution control factor, a manufacturing path control instruction parameter set is constructed. The manufacturing path control instruction includes:

[0128] Scan path structure parameters, including the three-dimensional coordinate sequence of the scanning segment, the scanning path direction vector and the scanning segment jump strategy;

[0129] Laser power control parameters, including roughness trend control coefficient and laser power base value and adjustment range under spatial distribution mapping;

[0130] Scanning speed control parameters, including a scanning speed function based on the roughness trend change rate, defining the scanning start speed, maximum speed, and speed gradient change form;

[0131] Scan spacing control parameters, including scan overlap, interlayer offset, and path spacing adjustment corresponding to changes in roughness spatial density;

[0132] The parameters in the manufacturing path control instruction parameter set are jointly adjusted according to the intersection area of ​​the roughness trend parameter and the spatial probability distribution characteristics of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber. A fusion control mapping model is constructed to generate manufacturing path control instruction sets for different manufacturing areas and time periods.

[0133] The manufacturing path control instruction set is converted into a manufacturing control file format that can be recognized by the aerospace propulsion chamber flow channel additive manufacturing equipment, and transmitted to the control system to complete the manufacturing execution.

[0134] In this implementation, the roughness trend parameter is expanded into a dynamic control factor with a time index. The spatial probability distribution characteristics of the inner wall roughness of the flow channel of the aerospace propulsion chamber are parsed into multiple sub-distribution clusters. Their spatial centers, distribution forms, and weight parameters are extracted as spatial control factors. A fusion control mapping model is constructed, and the temporal and spatial control factors are mapped to manufacturing path control parameters such as the scanning path, laser power, scanning speed, and spacing. This forms a manufacturing path control instruction set and converts it into a control file executable by the additive manufacturing equipment, achieving precise control of the manufacturing process. This method enables the manufacturing path to respond to the roughness evolution trend and spatial distribution changes in real time, thereby improving the surface consistency and forming accuracy of key inner wall structures and laying a foundation for high-quality manufacturing for thermal-fluid coupled enhancement design.

[0135] Example 1:

[0136] To verify the feasibility of this invention, the research team applied it to a propulsion chamber structural optimization project conducted at the combustion system experimental base of a certain aerospace research institute. The research team selected the combustion chamber of a small liquid oxygen-kerosene rocket engine as the verification object. They investigated the forming quality and surface roughness consistency of the internal cooling flow channels, and attempted to apply the proposed "manufacturing path control method based on the fusion of roughness trend and spatial distribution modeling" to the additive manufacturing process. The internal cooling channels of these engine combustion chambers are mostly spirally wound, with cross-sectional dimensions ranging from 0.8 mm to 1.5 mm. These channels have long paths and frequent spatial cross-sectional variations, and place extremely high demands on the wall roughness evolution stability, overall consistency, and defect control.

[0137] During the implementation process, a manufacturing parameter disturbance combination matrix including scanning speed, power, spacing and layer thickness disturbances was first established. Ten groups of channel structure samples were manufactured on IN718 powder material using metal laser selective melting additive manufacturing equipment. The roughness generation behavior under each combination of parameters was recorded. Then, a grey prediction model GM(1,1) was constructed. The trend characteristics of the channel roughness formation sequence were extracted by segmented accumulation and non-uniformly weighted background modeling. The roughness evolution behavior under different disturbance intervals was predicted, and the corresponding trend parameters and prediction sequences were generated. Three types of roughness indicators were collected from multiple areas of the sample channel, including the inlet section, middle section and outlet section. The three-dimensional spatial probability distribution model was constructed by combining the grey model trend parameters to guide the Dirichlet process, and a priori spatial model of the roughness multi-peak structure was established.

[0138] During the actual additive manufacturing process, confocal laser online measurement equipment is deployed to collect roughness metrics of the channel area during the build process in real time and calculate the time-point residuals between the residuals and the roughness trend prediction series. In the residual feedback mechanism, these residuals are constructed into residual feedback vectors using a sliding window approach. These are then introduced into the grey prediction model and Dirichlet modeling, respectively, to dynamically adjust the prediction accuracy and spatial modeling weights, achieving closed-loop coupled control between the roughness evolution model and the actual manufacturing state.

[0139] The spatiotemporal information extracted from the roughness trend parameters and spatial probability distribution parameters is converted into manufacturing path control instructions. The control strategy includes: generating three-dimensional scanning path vectors and jump structures, constructing a laser power base value adjustment function, defining a speed function and an overlap rate control model, and ultimately generating a set of control instruction files that take into account both regional roughness density and evolution trends and importing them into the equipment system for execution.

[0140] In terms of specific effects, the team conducted comparative experiments on 10 channel samples manufactured using the traditional strategy and the method of the present invention, measuring and recording the distribution of three types of roughness indicators on the inner wall surface of the channel, the Rq stability and defect rate of the middle section. Sample numbers A1-A10 correspond to samples of the traditional and present invention manufacturing strategies respectively.

[0141] Table 1 Comparative experimental results of roughness distribution of propulsion chamber channel

[0142]

[0143] Analysis of the data in the table "Comparative Experimental Results of Propulsion Chamber Channel Roughness Distribution" reveals that samples manufactured using the inventive method significantly outperform conventional methods in several key roughness metrics. First, in terms of arithmetic mean roughness (Ra), the average values ​​for conventional samples A1–A5 ranged from 15.9μm to 17.1μm, while those for samples A6–A10 using the inventive method significantly decreased to 13.1μm to 13.8μm, indicating a lower and smoother overall roughness level. Second, the coefficient of variation (R) decreased from 0.18–0.22 for the conventional method to 0.07–0.09, demonstrating that the new method provides more stable and consistent roughness control during the manufacturing process. Regarding maximum profile height (Rz), the maximum values ​​for conventional samples generally exceeded 80μm, while those using the inventive method maintained a stable Rz of 65–68μm, effectively suppressing localized roughness mutations. The stability index of the root mean square roughness (Rq) in the mid-section region increased from approximately 0.76 using the conventional method to 0.91 using the present invention, demonstrating that this method can significantly improve the morphological stability of the mid-section of the channel. The manufacturing defect rate, while averaging 7.2% using the conventional method, was significantly reduced to approximately 2.5% using the present invention's control strategy, demonstrating the present invention's ability to effectively suppress surface anomalies and structural defects. Comprehensive analysis demonstrates that the present invention's method achieves substantial improvements in roughness control accuracy, distribution consistency, and defect suppression, validating the significant effectiveness of the described manufacturing path control strategy in the additive manufacturing of complex channels.

[0144] In summary, the roughness trend prediction and spatial modeling fusion path control method proposed in the present invention can effectively improve the manufacturing consistency and surface quality of aerospace propulsion chamber channels, achieve stable control of roughness under multi-position and multi-scale conditions, and significantly reduce the incidence of manufacturing defects. It has high scalability and engineering practical value.

[0145] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A method for optimizing the roughness of a flow channel in additive manufacturing of an aerospace propulsion chamber, characterized in that: The following steps are involved: Set the manufacturing parameter perturbation strategy, perform additive manufacturing experiments under multiple perturbation groups, and collect the corresponding channel inner wall roughness formation sequence; Constructing a grey prediction model based on the collected roughness formation sequence , a roughness time evolution model is generated in the form of differential equations, in which the roughness changes with the disturbance conditions, and the roughness trend parameters and roughness prediction value sequence are output; The roughness distribution data at different locations on the channel wall are collected and combined with the roughness trend parameter as a priori input to guide the Dirichlet process to establish a spatial probability distribution model. Establish a fusion mechanism between roughness trend parameters and spatial probability distribution model, and output roughness trend parameters and roughness spatial probability distribution characteristics; Collect real-time roughness monitoring data during the additive manufacturing process, compare the real-time roughness monitoring data with the roughness prediction value sequence, and calculate the prediction residual; A residual feedback mechanism is constructed based on the prediction residual, and the residual feedback mechanism is introduced into the fusion mechanism to drive the linkage optimization process; Manufacturing path control instructions are generated based on the roughness trend parameters and roughness spatial probability distribution characteristics output by the fusion mechanism.

2. The method for optimizing the roughness of a flow channel in additive manufacturing of an aerospace propulsion chamber according to claim 1, characterized in that: The steps for setting the manufacturing parameter perturbation strategy, performing the additive manufacturing experiment, and collecting the channel inner wall roughness formation sequence include: Selecting laser power, scanning speed, scanning spacing and forming layer thickness as manufacturing parameter perturbation dimensions, constructing a manufacturing parameter perturbation combination matrix, and the manufacturing parameter perturbation combination matrix serves as a component of the manufacturing parameter perturbation strategy; Based on the manufacturing parameter perturbation combination matrix, additive forming experiments are sequentially performed on a metal laser selective melting additive manufacturing device, wherein the additive forming experiments generate samples including a flow channel structure of an aerospace propulsion chamber for different parameter combinations; After completing each set of additive manufacturing experiments, surface topography measurement equipment was used to perform multi-position roughness measurements on the inner wall of the flow channel of the aerospace propulsion chamber of the sample. The roughness measurement area covered multiple characteristic locations of the inlet section, midsection, and outlet section. The manufacturing parameter perturbation combination used in each set of additive forming experiments was paired with the corresponding roughness measurement results of the inner wall of the aerospace propulsion chamber flow channel to construct a roughness formation sequence.

3. The method for optimizing the roughness of a flow channel in additive manufacturing of an aerospace propulsion chamber according to claim 1, characterized in that: The grey prediction model is constructed based on the collected roughness formation sequence , a roughness time evolution model is generated in the form of a differential equation, in which the roughness changes with the disturbance conditions, and the roughness trend parameters and roughness prediction value sequence are output. Specifically, the following are included: The roughness formation sequence is constructed as the original time series data of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber. Based on the preset time window division rule, a segmented adjustable accumulation generation operation is performed to form multiple roughness gray generation subsequences. Based on each roughness grey generation subsequence, the corresponding grey prediction submodel is established respectively , and in each grey prediction sub-model Generate a set of roughness trend characteristic parameters for trend analysis; In each grey prediction sub-model In , a non-uniformly weighted background value sequence is constructed based on the roughness gray generator sequence: ； in, For the Item non-uniformly weighted background value sequence, Generate subsequences for roughness gray, is the weight coefficient, and its value range is , is the sequence index; Construct each grey prediction sub-model based on the non-uniformly weighted background value sequence The differential prediction equation of Each grey prediction sub-model The output roughness prediction value sequence is summarized; Each grey prediction sub-model The extracted roughness trend characteristic parameters are jointly integrated to form the roughness trend parameters.

4. The method for optimizing the roughness of a flow channel in additive manufacturing of an aerospace propulsion chamber according to claim 1, characterized in that: The method of collecting roughness distribution data at different locations on the channel wall and combining the roughness trend parameter as a priori input to guide the Dirichlet process to establish a spatial probability distribution model specifically includes: The inner wall of the flow channel in the aerospace propulsion chamber is divided into multiple spatial measurement areas. Three types of roughness indicators, namely arithmetic mean roughness, maximum profile height, and root mean square roughness, are collected in each spatial measurement area. The three-dimensional spatial position coordinates corresponding to each set of roughness indicators are recorded to construct the roughness distribution data of the inner wall of the flow channel in the aerospace propulsion chamber. The roughness distribution data of the inner wall of the flow channel of the aerospace propulsion chamber is used as the input of the non-parametric Bayesian model. The Dirichlet process adopts the multi-dimensional Gaussian distribution as the base distribution, and the concentration parameter is set to control the non-parametric prior strength of the number of clusters. In the Dirichlet process modeling, the initial state is an empty cluster. During the sampling process, each new data point takes the current three-dimensional spatial position coordinates and roughness index as input, and calculates the likelihood value of the distribution parameters of each existing cluster. The distribution parameters of the cluster include the mean vector and covariance matrix of the roughness index. According to the distribution parameters and likelihood value of each cluster and the corresponding cluster size, the concentration parameter is combined to calculate the belonging probability, perform cluster belonging sampling, and determine whether to generate a new roughness sub-distribution cluster; In each sampling round, the grey prediction model is introduced The output roughness trend parameter is used as a priori adjustment variable to update the cluster distribution weight so that the roughness evolution trend and spatial distribution structure are adjusted synchronously; After sampling convergence, all roughness sub-distribution clusters are merged to output the spatial probability distribution model of the inner wall roughness of the flow channel of the aerospace propulsion chamber.

5. The method for optimizing the roughness of a flow channel in additive manufacturing of a space propulsion chamber according to claim 1, characterized in that: The fusion mechanism for establishing the roughness trend parameter and the spatial probability distribution model specifically includes establishing a mapping relationship between the roughness trend parameter and each roughness sub-distribution cluster in the spatial probability distribution model in a one-to-one correspondence, calculating the residual between the roughness trend parameter and the mean of the corresponding roughness sub-distribution cluster in each mapping pair, and constructing a fusion weight adjustment function based on the residual, dynamically correcting the probability weight of each roughness sub-distribution cluster in the spatial probability distribution model of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber, and iteratively updating the roughness trend parameter using the residual as a feedback variable.

6. The method for optimizing the roughness of a flow channel in additive manufacturing of a space propulsion chamber according to claim 1, characterized in that: The collecting of real-time roughness monitoring data during the additive manufacturing process, comparing the real-time roughness monitoring data with a roughness prediction value sequence, and calculating the prediction residual specifically include: During the additive manufacturing process of the flow channel in aerospace propulsion chambers, a confocal laser measurement device was used to measure the inner wall of the flow channel online based on a preset scanning path and time interval. Three types of roughness indicators, including arithmetic mean roughness, maximum profile height, and root mean square roughness, were collected, along with the sampling time index and three-dimensional spatial position coordinates, forming real-time roughness monitoring data for the inner wall of the flow channel in aerospace propulsion chambers. The real-time roughness monitoring data of the inner wall of the flow channel of the aerospace propulsion chamber is reorganized into a time series structure according to the time index, and a one-to-one correspondence is made with the roughness prediction value sequence of the inner wall of the flow channel of the aerospace propulsion chamber in chronological order. The three types of predicted roughness indicators are matched with the three types of measured roughness indicators at each sampling moment. At each sampling time point, the numerical differences between the predicted values ​​and the measured values ​​of the three types of roughness, namely, arithmetic mean roughness, maximum profile height and root mean square roughness, were calculated as the prediction residuals at the corresponding time point.

7. The method for optimizing the roughness of a flow channel in additive manufacturing of a space propulsion chamber according to claim 1, characterized in that: The process of constructing a residual feedback mechanism based on prediction residuals and introducing the residual feedback mechanism into the fusion mechanism to drive the linkage optimization process specifically includes: Based on the real-time roughness monitoring data and the roughness prediction value series of the inner wall of the flow channel of the aerospace propulsion chamber, the roughness prediction values ​​and the real-time roughness monitoring values ​​are paired point by point according to the time index. The arithmetic mean roughness residual, maximum profile height residual and root mean square roughness residual at each time point are calculated, and a residual time series is generated. Segment the residual time series using a sliding window, calculate the mean, standard deviation, and trend slope of the residual series in each window, and construct a residual feedback vector; The residual feedback vector is used as the input of the residual feedback mechanism, acting on the roughness trend parameter generation process and the iterative process of the roughness spatial probability distribution model of the inner wall of the flow channel of the aerospace propulsion chamber, forming a dual-channel regulation mechanism. In the process of generating roughness trend parameters, the residual feedback mechanism is used to adjust the grey prediction model. The weight distribution structure and iterative update step size of model parameters in the construction method of non-uniformly weighted background values; In the spatial probability distribution model of the inner wall roughness of the flow channel in aerospace propulsion chambers, the residual feedback mechanism is used to adjust the sampling and new cluster generation thresholds of each roughness sub-distribution cluster, and dynamically correct the cluster distribution weights of the spatial distribution.

8. The method for optimizing the roughness of a flow channel in additive manufacturing of a space propulsion chamber according to claim 1, characterized in that: The manufacturing path control instructions specifically include: The roughness trend parameters are expanded by time index, and the roughness change rate, fluctuation amplitude and local evolution trend corresponding to different manufacturing time periods are extracted to form the time dynamic control factor in the manufacturing path control instruction; The spatial probability distribution characteristics of the roughness of the inner wall of the flow channel of the aerospace propulsion chamber are regionally decomposed, and the spatial position center, covariance matrix and roughness distribution weight of each roughness sub-distribution cluster are extracted to form the spatial distribution control factor in the manufacturing path control instruction. Based on the time dynamic control factor and the spatial distribution control factor, a manufacturing path control instruction is constructed. The manufacturing path control instruction includes: Scan path structure parameters, including the three-dimensional coordinate sequence of the scanning segment, the scanning path direction vector and the scanning segment jump strategy; Laser power control parameters, including roughness trend control coefficient and laser power base value and adjustment range under spatial distribution mapping; Scanning speed control parameters, including a scanning speed function based on the roughness trend change rate, defining the scanning start speed, maximum speed, and speed gradient change form; Scanning spacing control parameters include scanning overlap rate, interlayer offset and path spacing adjustment corresponding to the change of roughness spatial density.

Images