Real-time optimization method of fiber path in additive manufacturing of metal matrix fiber composites
Through enhanced learning model and hybrid optimization strategy, the fiber path of metal-based fiber composite materials is optimized in real time, and the problem of relying on experience and computing resources in the existing technology is solved, and efficient and accurate fiber path optimization and production process are achieved. Monitoring improves production efficiency and reduces manufacturing costs.
Patent Information
- Application Number
- CN202410955156.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-17
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2044-07-17
AI Technical Summary
In the additive manufacturing of existing metal-based fiber composite materials, fiber path design relies on experience and trial and error, and cannot be optimized in real time, resulting in waste of material performance, increased manufacturing cost and time. The existing real-time optimization method is single and computationally intensive, and it is not involved. Material preparation, model slicing, performance simulation analysis and post-processing.
Adopting enhanced learning model and hybrid optimization strategy, the initial fiber path is planned through the slice algorithm, combined with virtual simulation and physical testing for two-way verification, dynamically set optimization goals and constraints, iteratively optimize fiber paths until the expected goals are met, and real-time monitoring and manufacturing is monitored. Process, complete product manufacturing and testing.
Improve the production efficiency of metal-based fiber composite materials, shorten the product design to manufacturing cycle, ensure the accuracy and reliability of optimization results, reduce production risks and manufacturing costs, and provide comprehensiveness and system of the additive manufacturing process. sex.
Smart Images

Figure CN118504425B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of additive manufacturing, and in particular to a real-time optimization method for fiber paths in additive manufacturing of metal-based fiber composite materials. Background Art
[0002] With the continuous advancement of science and technology and the increasing engineering needs, metal-based fiber composites have been widely used in aerospace, automobile manufacturing, electronic equipment, and energy fields due to their high strength, high toughness, excellent wear resistance and corrosion resistance. However, the performance of metal-based fiber composites depends to a large extent on the layout and optimization of their fiber paths.
[0003] In the additive manufacturing process of metal-based fiber composites, the layout and design of the fiber path directly affect the final performance of the composite material. Traditional fiber path design methods often rely on experience and trial and error, and cannot be optimized in real time according to the actual situation in the manufacturing process, which not only leads to waste of material performance, but also increases manufacturing costs and time.
[0004] In recent years, with the rapid development of computer technology and optimization algorithms, real-time optimization technology has been gradually applied to the additive manufacturing of metal-based fiber composites. By collecting data from the manufacturing process in real time and using advanced optimization algorithms to adjust and optimize the fiber path in real time, the performance of the composites can be improved and the manufacturing cost and time can be reduced. However, the existing real-time optimization methods still have some shortcomings. On the one hand, due to the complexity and diversity of metal-based fiber composites, a single optimization algorithm is often difficult to meet all requirements; on the other hand, the real-time optimization process requires a lot of computing resources and has high requirements for computer hardware and software.
[0005] For example, a Chinese patent with authorization announcement number CN114147958B discloses a continuous fiber reinforced composite material with a high fiber content and a 3D printing method thereof, including: starting from the actual working conditions of continuous fiber 3D printing, optimizing the existing continuous fiber 3D printing path, proposing a staggered compaction printing strategy, adjusting the printing starting position of adjacent printed layers, so that the fibers of the next layer are laid on the resin bonding point of the previous layer, achieving a staggered compaction printing effect, and reducing the Z-axis lift when printing the next layer, compacting the continuous fibers in the upper layer of resin, thereby increasing the fiber volume content and increasing the interlayer bonding strength of adjacent printed layers.
[0006] The above existing technologies all have the following problems: 1) They focus on the optimization of printing technology but do not involve the comprehensive management of the entire additive manufacturing process, such as material preparation, model slicing, performance simulation analysis, and post-processing and performance testing of products; 2) The optimization methods are relatively single; 3) There is no clear mention of how to verify and evaluate its effects. Summary of the invention
[0007] In view of the shortcomings of the prior art, the present invention proposes a real-time optimization method for fiber paths in additive manufacturing of metal-based fiber composite materials, prepares fiber composite materials, obtains a three-dimensional model of the target component, plans the initial fiber path through a slicing algorithm, and uses a reinforcement learning model to predict and verify performance. Based on real-time data and performance analysis, a hybrid optimization strategy is used to iteratively optimize the fiber path until the expected goal is met. The optimization results are imported into the additive manufacturing equipment, the manufacturing process is monitored in real time, and the product manufacturing and testing are completed. The present invention improves the production efficiency of metal-based fiber composite materials through real-time optimization and monitoring.
[0008] To achieve the above object, the present invention provides the following technical solutions:
[0009] Real-time optimization method of fiber path in additive manufacturing of metal matrix fiber composites, including:
[0010] Step S1: preparing a metal-based fiber composite material, obtaining a three-dimensional model of a target component, and setting process parameters for additive manufacturing;
[0011] Step S2: Based on the three-dimensional model and process parameters, the three-dimensional model is sliced layer by layer according to the slice thickness using a contour slicing algorithm, and an initial fiber path is planned for each slice layer. The reinforcement learning model is introduced to predict the initial fiber path, and the prediction results are bidirectionally verified by combining virtual simulation and physical testing, and the performance parameters of the fiber material are evaluated to obtain performance prediction results;
[0012] Step S3: acquiring the current fiber path data in real time, dynamically setting the optimization goal and constraint conditions according to the performance prediction results and real-time data, introducing a hybrid optimization strategy to iteratively optimize the current fiber path, and performing performance simulation analysis on the optimized fiber path, evaluating the optimization effect, and obtaining a new fiber path;
[0013] Step S4: set the expected target, compare and analyze the new fiber path with the initial fiber path and the optimized fiber path. If the optimization effect reaches the expected target, output the new fiber path data and performance prediction results. Otherwise, adjust the optimization target and constraint conditions according to the comparison and analysis results, and return to step S3 for a new round of optimization.
[0014] Step S5: The optimization results are imported into the additive manufacturing equipment. After the manufacturing is completed, the product is post-processed and performance tested, and the test data is saved.
[0015] Specifically, the specific steps of step S2 include:
[0016] S2.1: Set the slice thickness, load the 3D model data, and divide the 3D model layer by layer along the Z-axis according to the slice thickness according to the contour algorithm. Extract the contour data of each slice layer, and determine the starting and ending points of the fiber on each slice according to the unidirectional laying rule of the fiber. Use the improved path planning algorithm to find the shortest path connecting the starting and ending points to obtain the initial fiber path;
[0017] S2.2: Construct a reinforcement learning model, and input the initial fiber path, contour data and process parameters into the reinforcement learning model to obtain the prediction results of the mechanical and thermal properties of the initial fiber path;
[0018] S2.3: Set the maximum error value of the reinforcement learning model , use finite element analysis to perform virtual simulation verification on the results predicted by the reinforcement learning model, and perform physical tests at the same time to obtain a two-way verification result X. If , then the parameters of the reinforcement learning model are automatically adjusted, where Represents the error of the two-way verification result X;
[0019] S2.4: Based on the bidirectional verification results, evaluate the strength, stiffness, and thermal stability performance of the fiber material, and integrate the evaluation results with the predictive analysis data of the reinforcement learning model to obtain the current fiber path status data.
[0020] Specifically, the specific formula of the path planning algorithm in S2.1 is:
[0021]
[0022] Among them, u and v represent the fiber path nodes on the slice layer, represents the initial fiber path from the starting point to the node v on the jth slice layer, t represents the time change, q represents the set of nodes that have been visited before, represents the shortest path from the starting point to node u on the jth slice layer, Indicates the jth slice layer from node u to node v at t and The edge weight at the state, represents the penalty function on the j-th slice layer, Indicates the intrinsic strength of the fiber. Indicates the fiber laying method, represents the change in direction of the fiber path between nodes u and v, represents the edge weight parameter, Represents the minimum function.
[0023] Specifically, the specific steps of step S3 include:
[0024] S3.1: Deploy a data acquisition system to capture the progress data of the manufacturing process and the status data of the current fiber path in real time, obtain a real-time data set, and pre-process the real-time data set;
[0025] S3.2: Train a big data analysis model, input the preprocessed real-time data set into the big data analysis model, perform data analysis based on the output results of the big data analysis model, and dynamically set optimization goals and constraints based on the data analysis results and real-time data;
[0026] S3.3: Based on the set optimization objectives and constraints, the current fiber path is iteratively optimized using a hybrid optimization strategy. During each iteration, the performance of the optimized fiber path is evaluated. , where r represents the number of iterations and is compared with the performance results of the previous iteration;
[0027] like , then adjust the optimization target weights and constraints according to the simulation results, and re-perform iterative optimization, where: Indicates Performance of the suboptimal fiber path;
[0028] like , the optimized fiber path is output as the new fiber path.
[0029] Specifically, the optimization objectives and constraints in S3.2 are:
[0030]
[0031] in, represents the minimum manufacturing cost of additive manufacturing, represents the cost of the i-th fiber material in the additive manufacturing process, represents the labor cost in the additive manufacturing process, represents the energy cost consumed in the additive manufacturing process, represents the transportation cost of the target component, represents the maximum production efficiency of the additive manufacturing process, represents the total production volume in the additive manufacturing process, represents the total duration of the additive manufacturing process, represents the defect rate of the target part in the additive manufacturing process, represents the number of target parts with defects in the additive manufacturing process, represents the minimum energy consumption of the additive manufacturing process, represents the energy consumed by the i-th fiber material in the additive manufacturing process, represents the minimum length of the fiber path during additive manufacturing, represents the fiber path length of the jth slice layer, represents the production time of the kth target component, represents the maximum production time allowed for the production of the kth target component, represents the capacity of the lth device, Indicates the maximum capacity of the lth device, represents the availability of the i-th fiber material in the additive manufacturing process, represents the quantity of the i-th fiber material required in the additive manufacturing process, represents the strength of the kth target component, represents the minimum strength of the kth target component, n represents the type of fiber material required for additive manufacturing, and m represents the number of slice layers for segmentation of the 3D model.
[0032] Specifically, the specific steps of the hybrid optimization strategy in S3.3 include:
[0033] S3.31: Generating an Initial Solution Set Using Random Methods , determine whether any solution in the initial solution set satisfies the constraint conditions. If not, regenerate it. An initial solution set represents a fiber path, where represents the x-th fiber path solution, where x represents the number of fiber path solutions;
[0034] S3.32: Set the optimization target weight and use the multi-objective optimization algorithm to generate the target function f. The formula is:
[0035] ;
[0036] S3.33: Use the objective function f to evaluate any solution in the initial solution set and calculate the fitness value of any solution. The formula is:
[0037]
[0038] in, represents the fitness value of the fiber path solution in the objective function f, Indicates The weight of the optimization objective.
[0039] Specifically, the specific steps of the hybrid optimization strategy in S3.3 also include:
[0040] S3.34: Set the fiber path solution size and initialize it. For each fiber path solution, the fitness value of the fiber path solution is calculated. , generate a new fiber path solution set;
[0041] S3.35: Assessment location The fitness value of , j represents the index of the slice layer
[0042] like , then update the location , get the updated position ;
[0043] like , then update the location , get the updated position ;
[0044] S3.36: Repeat steps S3.34 to S3.35 until the maximum number of iterations r is reached, and output As the final fiber path solution set.
[0045] Specifically, the new fiber path solution generation formula in S3.34 is:
[0046]
[0047] in, represents the xth fiber path solution on the jth slice layer. The speed of generation, represents the speed of the x-th fiber path solution at the j-th slice layer at the r-th generation, represents the xth fiber path solution on the jth slice layer. The position of the generation, represents the position of the x-th fiber path solution at the r-th generation on the j-th slice layer, represents the inertia weight, represents the optimal position of the x-th fiber path solution on the j-th slice layer, represents the optimal position of the fiber path solution on the jth slice layer, represents the position of the o-th neighbor fiber path solution on the j-th slice layer, represents the speed change rate weight, represents the velocity change of the x-th fiber path solution on the j-th slice layer, represents the noise weight, represents the noise term of the x-th fiber path solution on the j-th slice layer, represents the learning factor, represents a random number, and , N represents the number of neighbor fiber path solutions, and o represents the position index of the neighbor fiber path solution.
[0048] An electronic device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, steps of a method for real-time optimization of fiber paths in additive manufacturing of metal-based fiber composite materials are implemented.
[0049] A computer-readable storage medium stores computer instructions, which, when executed, execute the steps of a method for real-time optimization of fiber paths in additive manufacturing of metal-based fiber composite materials.
[0050] Compared with the prior art, the present invention has the following beneficial effects:
[0051] 1. The present invention proposes a real-time optimization method for fiber paths in additive manufacturing of metal-based fiber composite materials, introduces a reinforcement learning model and a hybrid optimization strategy, accelerates the optimization speed of the fiber path, and shortens the cycle from product design to actual manufacturing; the prediction results are bidirectionally verified by virtual simulation and physical testing to ensure the accuracy and reliability of the optimization results, and the application of a real-time monitoring system can timely discover and solve problems in the production process, thereby reducing production risks and additive manufacturing costs.
[0052] 2. The present invention proposes a real-time optimization method for fiber paths in additive manufacturing of metal-based fiber composite materials, which provides a complete process from material preparation to product post-processing and performance testing, ensuring the comprehensiveness and systematicness of the 3D printing process; dynamically setting optimization goals and constraints based on performance analysis results and real-time data, improving system flexibility and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0053] Figure 1 A schematic diagram of a real-time optimization method for fiber paths in additive manufacturing of metal-based fiber composite materials according to the present invention;
[0054] Figure 2 This is a workflow diagram of the fiber path real-time optimization method in additive manufacturing of metal-based fiber composite materials of the present invention;
[0055] Figure 3 A fiber path construction flow chart of the fiber path real-time optimization method in additive manufacturing of metal-based fiber composite materials of the present invention;
[0056] Figure 4 This is a fiber path optimization flow chart of the fiber path real-time optimization method in additive manufacturing of metal-based fiber composite materials of the present invention. DETAILED DESCRIPTION
[0057] Example 1
[0058] See also Figure 1-Figure 2 , an embodiment provided by the present invention: a real-time optimization method for fiber path in additive manufacturing of metal-based fiber composite materials, comprising the following steps:
[0059] Step S1: preparing a metal-based fiber composite material, obtaining a three-dimensional model of a target component, and setting process parameters for additive manufacturing;
[0060] Furthermore, the specific steps include: preparing a metal matrix and a fiber material, pretreating the fiber material, surface treating the metal matrix, mixing the pretreated fiber and the metal matrix according to a proportion and arrangement to obtain a metal-based fiber composite material, obtaining a three-dimensional model of the target component, and setting the process parameters for additive manufacturing.
[0061] The target part refers to the specific physical object or component that is ultimately desired to be produced through additive manufacturing technology. It can be a part, a structure, a tool, a part of a device, or any item that needs to be manufactured using metal matrix fiber composite materials through additive manufacturing technology.
[0062] Process parameters include: slice thickness, printing speed, and temperature.
[0063] Step S2: Based on the three-dimensional model and process parameters, the three-dimensional model is sliced layer by layer according to the slice thickness using a contour slicing algorithm, and an initial fiber path is planned for each slice layer. The reinforcement learning model is introduced to predict the initial fiber path, and the prediction results are bidirectionally verified by combining virtual simulation and physical testing, and the performance parameters of the fiber material are evaluated to obtain performance prediction results;
[0064] Step S3: acquiring the current fiber path data in real time, dynamically setting the optimization goal and constraint conditions according to the performance prediction results and real-time data, introducing a hybrid optimization strategy to iteratively optimize the current fiber path, and performing performance simulation analysis on the optimized fiber path, evaluating the optimization effect, and obtaining a new fiber path;
[0065] The hybrid optimization strategy adopted in the present invention is to iteratively optimize the current fiber path by combining an improved multi-objective optimization algorithm with an improved particle swarm optimization strategy, thereby improving the global search capability, the quality and efficiency of the path solution, the real-time and dynamic performance, and the path optimization effect.
[0066] Step S4: set the expected target, compare and analyze the new fiber path with the initial fiber path and the optimized fiber path. If the optimization effect reaches the expected target, output the new fiber path data and performance prediction results. Otherwise, adjust the optimization target and constraint conditions according to the comparison and analysis results, and return to step S3 for a new round of optimization.
[0067] Step S5: Import the optimization results into the additive manufacturing equipment, combine with the real-time monitoring system, and carry out actual manufacturing. The actual manufacturing data is collected through the real-time monitoring system and compared with the predicted data. After the manufacturing is completed, the product is post-processed and performance tested, and the actual test data is saved.
[0068] Example 2
[0069] See also Figure 3 In this embodiment, the specific steps of step S2 include:
[0070] S2.1: Set the slice thickness, load the 3D model data, and divide the 3D model layer by layer along the Z-axis according to the slice thickness according to the contour algorithm. Extract the contour data of each slice layer, and determine the starting and ending points of the fiber on each slice according to the unidirectional laying rule of the fiber. Use the improved path planning algorithm to find the shortest path connecting the starting and ending points to obtain the initial fiber path;
[0071] The specific steps of segmentation using the contour algorithm include:
[0072] (1) Obtain the 3D model to be processed, import the 3D model into the slicing software, and set the slicing parameters;
[0073] (2) The slicing software will slice the 3D model layer by layer along the Z-axis direction starting from the bottom of the model according to the set slice thickness. Each slice layer is a two-dimensional plane, representing the cross-sectional shape of the 3D model at that height.
[0074] (3) For each slice layer, all closed outer boundaries and inner boundaries on the slice layer are found through a contour detection algorithm to form contour line data, wherein the contour detection algorithm is a prior art in the art and is not an inventive solution of the present application, and is not described in detail here;
[0075] (4) Clean the extracted contour data to remove noise, overlap or unnecessary parts;
[0076] (5) Optimize the cleaned contour data, such as simplifying the shape and smoothing the curve, and save the processed contour data in formats including G-code and SVG;
[0077] (6) Export all processed slice layer data to form a file or file set containing all slice information, and check the exported slice data to ensure the integrity and accuracy of the data;
[0078] (7) Send the slice data to the additive manufacturing equipment for the actual manufacturing process.
[0079] S2.2: Construct a reinforcement learning model, and input the initial fiber path, contour data and process parameters into the reinforcement learning model to obtain the prediction results of the mechanical and thermal properties of the initial fiber path;
[0080] The specific steps to build a reinforcement learning model to predict performance include:
[0081] (1) Clarify the problem and define the goal, define the state, action, and reward. The problem is to predict the mechanical and thermal properties of the initial fiber path;
[0082] (2) Selecting the Q-learning algorithm as the reinforcement learning algorithm, wherein the Q-learning algorithm is the prior art content in this field and is not the inventive solution of the present application, and is not described in detail here;
[0083] (3) constructing a multi-neural network model and randomly initializing the weights and biases of the neural network, wherein the construction of the neural network model is the prior art content in this field and is not the inventive solution of the present application, and is not described in detail here;
[0084] (4) collecting or generating a series of initial fiber paths as input data, while collecting contour data and process parameters, labels or reward functions;
[0085] (5) Let the agent interact with the environment according to the current strategy, collect experience data, store the experience data in the replay buffer, sample from the experience data and update the weights of the neural network to minimize the prediction error or maximize the expected reward;
[0086] (6) Use the validation set or test set to evaluate the performance of the multi-neural network and adjust the hyperparameters based on the evaluation results;
[0087] (7) The trained multi-neural network is used to predict the mechanical and thermal properties of the new initial fiber path, and the fiber path or process parameters are adjusted according to the prediction results to optimize the performance.
[0088] S2.3: Set the maximum error value of the reinforcement learning model , use finite element analysis to perform virtual simulation verification on the results predicted by the reinforcement learning model, and perform physical tests at the same time to obtain a two-way verification result X. If , then the parameters of the reinforcement learning model are automatically adjusted, where Represents the error of the two-way verification result X;
[0089] Furthermore, the specific steps of verifying the prediction results of the reinforcement learning model by finite element analysis include:
[0090] (1) Collect initial fiber path, contour data and process parameters, as well as relevant material properties and boundary condition information;
[0091] (2) Establish a finite element model in the finite element analysis software and mesh the finite element model to divide the model into small units;
[0092] (3) Define material properties for each unit or region in the finite element model based on process parameters and actual material properties, such as elastic modulus, Poisson's ratio, and thermal conductivity. Apply boundary conditions, such as fixed constraints and symmetry constraints, to the finite element model based on actual conditions. At the same time, apply loads, where loads include mechanical loads and thermal loads. Mechanical loads include pressure, tension, and bending moment, and thermal loads include temperature gradients.
[0093] (4) Set the solver parameters and solution accuracy according to the complexity of the problem and the required accuracy;
[0094] (5) Run finite element analysis to calculate the mechanical and thermal responses of the model and monitor the solution process to ensure that the calculation is stable and convergent;
[0095] (6) Post-process and analyze the simulation results to extract performance indicators such as stress distribution, deformation, and temperature field;
[0096] (7) Compare the finite element analysis results with the prediction results of the reinforcement learning model to evaluate the prediction accuracy of the reinforcement learning model;
[0097] (8) If the prediction error of the reinforcement learning model is large, iterative optimization is performed based on the simulation results to adjust the model parameters or structure to improve the prediction accuracy;
[0098] (9) Conduct physical tests to verify the accuracy of the simulation results and compare them with the finite element analysis results.
[0099] S2.4: Based on the bidirectional verification results, evaluate the strength, stiffness, and thermal stability performance of the fiber material, and integrate the evaluation results with the predictive analysis data of the reinforcement learning model to obtain the current fiber path status data.
[0100] The specific formula of the path planning algorithm in S2.1 is:
[0101]
[0102] Among them, u and v represent the fiber path nodes on the slice layer, represents the initial fiber path from the starting point to the node v on the jth slice layer, t represents the time change, q represents the set of nodes that have been visited before, represents the shortest path from the starting point to node u on the jth slice layer, Indicates the jth slice layer from node u to node v at t and The edge weight at the state, represents the penalty function on the j-th slice layer, Indicates the intrinsic strength of the fiber. Indicates the fiber laying method, represents the change in direction of the fiber path between nodes u and v, represents the edge weight parameter, Represents the minimum function.
[0103] Example 3
[0104] See also Figure 4 In this embodiment, the specific steps of step S3 include:
[0105] S3.1: Deploy a data acquisition system to capture the progress data of the manufacturing process and the status data of the current fiber path in real time, obtain a real-time data set, and pre-process the real-time data set;
[0106] S3.2: Train a big data analysis model, input the preprocessed real-time data set into the big data analysis model, perform data analysis based on the output results of the big data analysis model, and dynamically set optimization goals and constraints based on the data analysis results and real-time data;
[0107] The specific steps for training a big data analysis model include:
[0108] (1) Collect real-time data sets, clean and preprocess the collected data, including data deduplication, missing value processing, data format conversion, outlier processing, and divide the preprocessed data sets into training sets, validation sets, and test sets;
[0109] (2) Use neural network algorithms and Spark tools to build big data analysis models and set model parameters, network structure, activation function, and loss function;
[0110] 1) Use Spark to read data from data sources, including HDFS, HBase, and Cassandra, and preprocess the data, including data cleaning, data type conversion, and data format normalization. Data cleaning includes removing invalid columns, missing values, and duplicates;
[0111] 2) Divide the data into training set, validation set and test set;
[0112] 3) Extract features from the preprocessed data, and scale and encode the extracted features;
[0113] 4) Select the recurrent neural network architecture, use the MLlib learning library in the Spark environment to build a recurrent neural network model, and set the parameters of the recurrent neural network model, such as learning rate, number of iterations, number of hidden layers, and number of neurons;
[0114] 5) Use Spark's distributed computing capabilities to train the recurrent neural network model, divide the training data into multiple partitions, and calculate the gradient in parallel on each partition. At the same time, use the stochastic gradient descent optimization algorithm to update the weights of the recurrent neural network model, and monitor the performance of the recurrent neural network model through the validation set;
[0115] 6) Use the test data set to evaluate the trained recurrent neural network model, calculate the performance indicators of the recurrent neural network model, such as accuracy, recall, and F1 score, analyze the performance of the model, and make adjustments based on the evaluation results.
[0116] (3) Use the training data set to train the big data analysis model. During the training process, the big data analysis model will learn to extract features from the input data and predict the output, and adjust the model parameters and structure based on the performance of the validation set to optimize the model performance;
[0117] (4) Use the test data set to evaluate the trained big data analysis model. The evaluation indicators include accuracy, recall, F1 score, and AUC.
[0118] (5) Analyze the performance of the big data analysis model and adjust or improve the big data analysis model based on the evaluation results;
[0119] (6) If the big data analysis model achieves satisfactory performance, it will be deployed to the production environment to predict and analyze real-time data. During the deployment process, it is necessary to ensure the effective integration of the model with the real-time data stream and to ensure the real-time performance and scalability of the model.
[0120] S3.3: Based on the set optimization objectives and constraints, the current fiber path is iteratively optimized using a hybrid optimization strategy. During each iteration, the performance of the optimized fiber path is evaluated. , where r represents the number of iterations and is compared with the performance results of the previous iteration;
[0121] like , then adjust the optimization target weights and constraints according to the simulation results, and re-perform iterative optimization, where: Indicates Performance of the suboptimal fiber path;
[0122] like , the optimized fiber path is output as the new fiber path.
[0123] Specifically, the optimization objectives and constraints in S3.2 are:
[0124]
[0125] in, represents the minimum manufacturing cost of additive manufacturing, represents the cost of the i-th fiber material in the additive manufacturing process, represents the labor cost in the additive manufacturing process, represents the energy cost consumed in the additive manufacturing process, represents the transportation cost of the target component, represents the maximum production efficiency of the additive manufacturing process, represents the total production volume in the additive manufacturing process, represents the total duration of the additive manufacturing process, represents the defect rate of the target part in the additive manufacturing process, represents the number of target parts with defects in the additive manufacturing process, represents the minimum energy consumption of the additive manufacturing process, represents the energy consumed by the i-th fiber material in the additive manufacturing process, represents the minimum length of the fiber path during additive manufacturing, represents the fiber path length of the jth slice layer, represents the production time of the kth target component, represents the maximum production time allowed for the production of the kth target component, represents the capacity of the lth device, Indicates the maximum capacity of the lth device, represents the availability of the i-th fiber material in the additive manufacturing process, represents the quantity of the i-th fiber material required in the additive manufacturing process, represents the strength of the kth target component, represents the minimum strength of the kth target component, n represents the type of fiber material required for additive manufacturing, and m represents the number of slice layers for segmentation of the 3D model.
[0126] The specific steps of the hybrid optimization strategy in S3.3 include:
[0127] S3.31: Generating an Initial Solution Set Using Random Methods , determine whether any solution in the initial solution set satisfies the constraint conditions. If not, regenerate it. An initial solution set represents a fiber path, where represents the x-th fiber path solution, where x represents the number of fiber path solutions;
[0128] The specific steps of generating the initial solution set by random method include:
[0129] (1) Define the search space for fiber path optimization, that is, the set of all possible fiber paths, and determine the constraints of the optimization problem;
[0130] (2) Initialize a random seed;
[0131] (3) In the search space, the control points, directions, and lengths of the fiber paths are randomly sampled, and multiple initial fiber paths are randomly generated as the initial solution set according to the constraints;
[0132] (4) evaluating each fiber path in the initial solution set using a predefined performance metric;
[0133] (5) Starting from the initial solution set, an iterative optimization strategy is used.
[0134] S3.32: Set the optimization target weight and use the multi-objective optimization algorithm to generate the target function f. The formula is:
[0135] ;
[0136] S3.33: Use the objective function f to evaluate any solution in the initial solution set and calculate the fitness value of any solution. The formula is:
[0137]
[0138] in, represents the fitness value of the fiber path solution in the objective function f, Indicates The weight of the optimization objective.
[0139] The specific steps of the hybrid optimization strategy in S3.3 also include:
[0140] S3.34: Set the fiber path solution size and initialize it. For each fiber path solution, the fitness value of the fiber path solution is calculated. , generate a new fiber path solution set;
[0141] S3.35: Assessment location The fitness value of ;
[0142] like , then update the location , get the updated position ;
[0143] like , then update the location , get the updated position ;
[0144] S3.36: Repeat steps S3.34 to S3.35 until the maximum number of iterations r is reached, and output As the final fiber path solution set.
[0145] The new fiber path solution generation formula in S3.34 is:
[0146]
[0147] in, represents the fiber path solution at the jth slice layer. The speed of generation, represents the speed of the x-th fiber path solution at the j-th slice layer at the r-th generation, represents the xth fiber path solution on the jth slice layer. The position of the generation, represents the position of the x-th fiber path solution at the r-th generation on the j-th slice layer, represents the inertia weight, represents the optimal position of the x-th fiber path solution on the j-th slice layer, represents the optimal position of the fiber path solution on the jth slice layer, represents the position of the o-th neighbor fiber path solution on the j-th slice layer, represents the speed change rate weight, represents the velocity change of the x-th fiber path solution on the j-th slice layer, represents the noise weight, represents the noise term of the x-th fiber path solution on the j-th slice layer, represents the learning factor, represents a random number, and , N represents the number of neighbor fiber path solutions, and o represents the position index of the neighbor fiber path solution.
[0148] Example 4
[0149] An electronic device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, steps of a method for real-time optimization of fiber paths in additive manufacturing of metal-based fiber composite materials are implemented.
[0150] A computer-readable storage medium stores computer instructions, which, when executed, execute the steps of a method for real-time optimization of fiber paths in additive manufacturing of metal-based fiber composite materials.
[0151] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in the field may also change, modify, replace and modify the above-mentioned embodiments without departing from the scope of protection of the purpose of the present invention and the claims, and all of these are within the protection of the present invention.
Claims
1. A real-time optimization method for fiber paths in additive manufacturing of metal-based fiber composite materials, characterized in that: include: Step S1: preparing a metal-based fiber composite material, obtaining a three-dimensional model of a target component, and setting process parameters for additive manufacturing; Step S2: Based on the three-dimensional model and process parameters, the three-dimensional model is sliced layer by layer according to the slice thickness using a contour slicing algorithm, and an initial fiber path is planned for each slice layer. The reinforcement learning model is introduced to predict the initial fiber path, and the prediction results are bidirectionally verified by combining virtual simulation and physical testing, and the performance parameters of the fiber material are evaluated to obtain performance prediction results; Step S3: acquiring the current fiber path data in real time, dynamically setting the optimization goal and constraint conditions according to the performance prediction results and real-time data, introducing a hybrid optimization strategy to iteratively optimize the current fiber path, and performing performance simulation analysis on the optimized fiber path, evaluating the optimization effect, and obtaining a new fiber path; Step S4: set the expected target, compare and analyze the new fiber path with the initial fiber path and the optimized fiber path. If the optimization effect reaches the expected target, output the new fiber path data and performance prediction results. Otherwise, adjust the optimization target and constraint conditions according to the comparison and analysis results, and return to step S3 for a new round of optimization. Step S5: The optimization results are imported into the additive manufacturing equipment. After the manufacturing is completed, the product is post-processed and performance tested, and the test data is saved.
2. The real-time optimization method for fiber paths in additive manufacturing of metal-based fiber composite materials according to claim 1, characterized in that: The specific steps of step S2 include: S2.1: Set the slice thickness, load the 3D model data, and divide the 3D model layer by layer along the Z-axis according to the slice thickness according to the contour algorithm. Extract the contour data of each slice layer, and determine the starting and ending points of the fiber on each slice according to the unidirectional laying rule of the fiber. Use the improved path planning algorithm to find the shortest path connecting the starting and ending points to obtain the initial fiber path; S2.2: Construct a reinforcement learning model, and input the initial fiber path, contour data and process parameters into the reinforcement learning model to obtain the prediction results of the mechanical and thermal properties of the initial fiber path; S2.3: Set the maximum error value of the reinforcement learning model , use finite element analysis to perform virtual simulation verification on the results predicted by the reinforcement learning model, and perform physical tests at the same time to obtain a two-way verification result X. If , then the parameters of the reinforcement learning model are automatically adjusted, where Represents the error of the two-way verification result X; S2.4: Based on the bidirectional verification results, evaluate the strength, stiffness, and thermal stability performance of the fiber material, and integrate the evaluation results with the predictive analysis data of the reinforcement learning model to obtain the current fiber path status data.
3. The real-time optimization method for fiber path in additive manufacturing of metal-based fiber composite materials according to claim 2, characterized in that: The specific formula of the path planning algorithm in S2.1 is: ; in, u , v represents the fiber path node on the slice layer, Indicates j From the starting point to the node v The initial fiber path, t represents the time variation, q Represents the set of nodes that have been visited before. Indicates j From the starting point to the node u The shortest path of Indicates j Slice layer slave node u To Node v exist t and The edge weight at the state, Indicates j The penalty function on the slice layer is Indicates the intrinsic strength of the fiber. Indicates the fiber laying method, Indicates the fiber path at the node u and v The change in direction between represents the edge weight parameter, Represents the minimum function.
4. The real-time optimization method for fiber path in additive manufacturing of metal-based fiber composite materials according to claim 3, characterized in that: The specific steps of step S3 include: S3.1: Deploy a data acquisition system to capture the progress data of the manufacturing process and the status data of the current fiber path in real time, obtain a real-time data set, and pre-process the real-time data set; S3.2: Train a big data analysis model, input the preprocessed real-time data set into the big data analysis model, perform data analysis based on the output results of the big data analysis model, and dynamically set optimization goals and constraints based on the data analysis results and real-time data; S3.3: Based on the set optimization objectives and constraints, the current fiber path is iteratively optimized using a hybrid optimization strategy. During each iteration, the performance of the optimized fiber path is evaluated. ,in, r Indicates the number of iterations and compares the performance results with the previous iteration; like , then adjust the optimization target weights and constraints according to the simulation results, and re-perform iterative optimization, where: Indicates Performance of the suboptimal fiber path; like , the optimized fiber path is output as the new fiber path.
5. The real-time optimization method for fiber path in additive manufacturing of metal-based fiber composite materials according to claim 4, characterized in that: The optimization objectives and constraints in S3.2 are: ; in, represents the minimum manufacturing cost of additive manufacturing, In the additive manufacturing process i The cost of fiber materials, represents the labor cost in the additive manufacturing process, represents the energy cost consumed in the additive manufacturing process, represents the transportation cost of the target component, represents the maximum production efficiency of the additive manufacturing process, represents the total production volume in the additive manufacturing process, represents the total duration of the additive manufacturing process, represents the defect rate of the target part in the additive manufacturing process, represents the number of target parts with defects in the additive manufacturing process, represents the minimum energy consumption of the additive manufacturing process, In the additive manufacturing process i Energy consumed by fiber materials, represents the minimum length of the fiber path during additive manufacturing, Indicates j The fiber path length of each slice layer, Indicates k The production time of each target component, Indicates k The maximum production time allowed for the production of a target component, Indicates l Equipment capacity, Indicates l The maximum capacity of a device, In the additive manufacturing process i The availability of fiber materials, Indicates the first i The number of fiber materials, Indicates k The strength of the target component, Indicates k The minimum strength of the target component, n Indicates the type of fiber material required for additive manufacturing, m Indicates the number of slice layers for 3D model segmentation.
6. The real-time optimization method for fiber path in additive manufacturing of metal-based fiber composite materials according to claim 5, characterized in that: The specific steps of the hybrid optimization strategy in S3.3 include: S3.31: Generating an Initial Solution Set Using Random Methods , determine whether any solution in the initial solution set satisfies the constraint condition. If not, regenerate it. An initial solution set represents a fiber path, where Indicates x fiber path solutions, x represents the number of fiber path solutions; S3.32: Set optimization target weights and generate target functions using multi-objective optimization algorithms f , the formula is: ; S3.33: Using the objective function f Evaluate any solution in the initial solution set and calculate the fitness value of any solution. The formula is: ; in, Represents the objective function f The fitness value of the fiber path solution in , Indicates The weight of the optimization objective.
7. The real-time optimization method for fiber path in additive manufacturing of metal-based fiber composite materials according to claim 6, characterized in that: The specific steps of the hybrid optimization strategy in S3.3 also include: S3.34: Set the fiber path solution size and initialize it. For each fiber path solution, the fitness value of the fiber path solution is calculated. , generate a new fiber path solution set; S3.35: Assessment location The fitness value of , j represents the index of the slice layer; like , then update the location , get the updated position ; like , then update the location , get the updated position ; S3.36: Repeat steps S3.34 to S3.35 until the maximum number of iterations is reached r , output As the final fiber path solution set.
8. The method for real-time optimization of fiber paths in additive manufacturing of metal-based fiber composite materials according to claim 7, characterized in that: The new fiber path solution generation formula in S3.34 is: ; in, Indicates x The fiber path solution is j Slice layer The speed of generation, Indicates x The fiber path solution is j Slice layer r The speed of generation, Indicates x The fiber path solution is j Slice layer The position of the generation, Indicates x The fiber path solution is j Slice layer r The position of the generation, represents the inertia weight, Indicates x The fiber path solution is j The optimal position on each slice layer, The fiber path solution is j The optimal position on each slice layer, Indicates o The neighbor fiber path solution is in j The position on the slice layer, represents the speed change rate weight, Indicates x The fiber path solution is j The velocity variation on each slice layer, represents the noise weight, Indicates x The fiber path solution is j The noise term on the slice layer, represents the learning factor, and , N represents the number of neighbor fiber path solutions, o represents the position index of the neighbor fiber path solution, represents a random number, and .
9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for real-time optimization of fiber paths in additive manufacturing of metal-based fiber composite materials according to any one of claims 1 to 8 are implemented.
10. A computer-readable storage medium, characterized in that: Computer instructions are stored thereon, and when the computer instructions are executed, the steps of the method for real-time optimization of fiber paths in additive manufacturing of metal-based fiber composite materials as described in any one of claims 1 to 8 are executed.
Citation Information
Patent Citations
A high-fiber-content continuous fiber reinforced composite material and its 3D printing method
CN114147958B
Complex mold cavity electric arc additive trajectory generation method
CN116532755A
3D printing method for continuous fiber composite path
CN116674195A
Shaping method by rapid prototyping and shaped article
JP2005074931A
Tool Path Generator with Embedded Process Control Commands for Additive Manufacturing
US20170057172A1
