A method and system for topology optimization of an additive manufactured part structure

Abstract

The application provides a kind of additive manufacturing part structure topology optimization method and system, it is related to additive manufacturing technical field, the method comprises: according to the working condition information of the part to be optimized to determine evaluation function, establish evaluation model;Network optimization model is constructed, the output information of the network optimization model is evaluated using the evaluation model, whether the network optimization model reaches preset reliability is judged;If the network optimization model reaches the preset reliability, the structure of the part to be optimized is optimized using the network optimization model, and an optimized structure is obtained.The technical problems of large amount of calculation and low calculation efficiency in the prior art topology optimization method are solved, the topology optimization result is more reliable, the efficiency of topology optimization finite element calculation is improved, the calculation speed is improved, and the cost is saved.

Description

Technical Field

[0001] This invention relates to the field of additive manufacturing technology, and in particular to a method and system for optimizing the topology of additively manufactured parts. Background Technology

[0002] In recent years, structural topology optimization methods and related algorithms have developed rapidly, finding applications not only in traditional mechanical design and manufacturing but also in high-end engineering manufacturing fields such as aerospace and marine engineering. Structural topology optimization aims to find the optimal material distribution to achieve certain performance requirements of the structure under given constraints. For lightweight structural design, it can effectively reduce weight and improve flexibility. Additive manufacturing, also known as 3D printing, differs from traditional subtractive manufacturing by constructing objects through layer-by-layer material stacking, allowing for topology optimization based on the most efficient force transmission path. Currently, optimization of structural parts is mainly based on the finite element method, which updates the material distribution within the design domain by analyzing the stress of unit structures. Without affecting structural safety performance, elements with low contribution to the structure are removed to achieve lightweight design. Major structural optimization methods include: Solid Isotropic Material Penalty Method (SIMP), Two-Way Progressive Structural Optimization Method (BESO), Movable Deformable Component Method (MMC), and level set methods. Although the above research results have significantly improved the computational efficiency and scale of topology optimization methods for solving problems, they are often limited by the computational nature of iterative topology optimization, which restricts the overall efficiency improvement of topology optimization methods. Summary of the Invention

[0003] In view of this, the first aspect of the present invention provides a method for topology optimization of additive manufacturing parts, to solve the technical problems of high computational load and low computational efficiency in existing topology optimization methods. The method includes:

[0004] Based on the working condition information of the part to be optimized, define the weight coefficients and constraints for different working conditions, and determine the evaluation function based on the weight coefficients and constraints.

[0005] An evaluation model is established based on the evaluation function;

[0006] Construct a network optimization model, use the evaluation model to evaluate the output information of the network optimization model, and determine whether the network optimization model has reached a preset reliability.

[0007] If the network optimization model reaches the preset reliability, the network optimization model is used to optimize the structure of the part to be optimized, and an optimized structure is obtained.

[0008] Furthermore, the construction of the network optimization model includes:

[0009] A finite element model of the part structure is established, and the part to be optimized is iterated a preset number of times using the finite element model of the part structure to obtain the element update status and performance data of the part to be optimized during the iteration process;

[0010] A training dataset is constructed based on the unit update status and the performance data;

[0011] The long short-term memory recurrent neural network is trained based on the training dataset to obtain the optimized network model. The optimized network model is then used to optimize the part to be optimized, thereby obtaining the output information. Further, training the long short-term memory recurrent neural network based on the training dataset to obtain the optimized network model includes:

[0012] During the training of the long short-term memory recurrent neural network based on the training dataset, the long short-term memory recurrent neural network is trained based on the backpropagation algorithm to obtain an optimized network model, wherein the long short-term memory recurrent neural network includes a first hidden layer and a second hidden layer.

[0013] Furthermore, both the first hidden layer and the second hidden layer have 25 nodes, and the network optimization model is optimized using the ReLU activation function, the mean absolute error (MAE), and the RMSPropOptimizer, respectively.

[0014] Furthermore, when training the long short-term memory recurrent neural network based on the training dataset, the training parameters are set to a learning rate of 0.001 and a sample size of 1000.

[0015] Furthermore, the evaluation function G(X,A,ρ)=V(X)+ρ×AR(X); AR(X)=∑a i ×r i ;

[0016] Where V(X) is the volume fraction, R(X) is the constraint condition, ρ is the penalty intensity, AR(X) is the constraint condition evaluation term under multiple working conditions, A is the weighting coefficient corresponding to different working conditions, and a i r is the weighting coefficient for the i-th working condition. i These are the constraints under the i-th operating condition.

[0017] A second aspect of this invention provides a topology optimization system for additive manufacturing parts, addressing the technical problems of high computational load and low computational efficiency in existing topology optimization methods. The system includes: an evaluation function acquisition module, used to define weight coefficients and constraints for different operating conditions based on the operating condition information of the part to be optimized, and to determine an evaluation function based on the weight coefficients and constraints;

[0018] An evaluation model construction module is used to build an evaluation model based on the evaluation function.

[0019] An optimization model construction module is used to construct a network optimization model, and to evaluate the output information of the network optimization model using the evaluation model to determine whether the network optimization model has reached a preset reliability.

[0020] The component structure optimization module is used to optimize the structure of the component to be optimized using the network optimization model if the network optimization model reaches the preset reliability, thereby obtaining an optimized structure.

[0021] Furthermore, the optimization model construction module is also used to establish a finite element model of the part structure, use the finite element model of the part structure to perform a preset number of iterations on the part to be optimized, and obtain the unit update status and performance data of the part to be optimized during the iteration process; construct a training dataset based on the unit update status and the performance data; train a long short-term memory recurrent neural network based on the training dataset to obtain the network optimization model, and use the network optimization model to optimize the part to be optimized to obtain the output information.

[0022] A third aspect of the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the above-described additive manufacturing part structure topology optimization methods.

[0023] A fourth aspect of the present invention also provides a computer-readable storage medium storing a computer program for executing any of the above-described additive manufacturing part structure topology optimization methods.

[0024] Compared with the prior art, the beneficial effects that can be achieved by at least one of the above-mentioned technical solutions adopted in this specification include: providing a topology optimization method for additive manufacturing parts, which can be used for topology optimization of complex parts under multiple working conditions by constructing a new evaluation function; and making the topology optimization results more reliable by applying a neural network optimization model to the process of tracking topology optimization, thereby improving the efficiency of finite element calculation for topology optimization, increasing the calculation speed, and saving costs. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a schematic diagram of a method for optimizing the topology of an additive manufacturing part structure provided in an embodiment of the present invention;

[0027] Figure 2 This is a schematic diagram of a topology optimization method for additive manufacturing parts provided in an embodiment of the present invention;

[0028] Figure 3 This invention provides a structural diagram of an additive manufacturing part topology optimization system.

[0029] Figure 4 This is a structural diagram of a computer device provided in an embodiment of the present invention.

[0030] In the figure, the following labels are used: 300, System; 310, Evaluation function acquisition module; 320, Evaluation model construction module; 330, Optimization model construction module; 340, Part structure optimization module; 401, Memory; 402, Processor. Detailed Implementation

[0031] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0032] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. This application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0033] In this embodiment of the invention, a method for topology optimization of additive manufacturing parts is provided. This method can be used for topology optimization of complex parts under multiple working conditions by constructing a new evaluation function. By applying a neural network optimization model to the process of tracking topology optimization, the topology optimization results are made more reliable, the efficiency of finite element calculation for topology optimization is improved, the calculation speed is increased, and the cost is saved.

[0034] like Figure 1 and Figure 2 As shown, the additive manufacturing part structure topology optimization method provided in this embodiment specifically includes the following steps:

[0035] Step S100: Define weight coefficients and constraints for different working conditions based on the working condition information of the part to be optimized, and determine the evaluation function based on the weight coefficients and constraints;

[0036] Specifically, based on the operating condition information of the part to be optimized, weight coefficients and constraints for different operating conditions are defined, and an evaluation function is established. The evaluation function is used to measure the performance and behavior of the model, and the choice of the evaluation function depends on the type of problem and the specific evaluation requirements.

[0037] In this embodiment, the evaluation function is defined as G(X,A,ρ).

[0038] Among them, G(X,A,ρ)=V(X)+ρ×AR(X); (1)

[0040] AR(X)=∑a i ×r i ; (2)

[0042] In formulas (1) and (2), V(X) is the volume fraction, R(X) is the constraint condition, ρ is the penalty intensity, AR(X) is the constraint condition evaluation term under multiple working conditions, A is the weighting coefficient corresponding to different working conditions, and a i r is the weighting coefficient for the i-th working condition. i Let A be the constraint condition for the i-th working condition. For example, if it is a single working condition, then A = 1, and R(X) is the objective function under the constraints of that working condition.

[0043] Step S200: Establish an evaluation model based on the evaluation function;

[0044] Specifically, model evaluation refers to a comprehensive performance assessment of a machine learning model. During the evaluation process, a predefined evaluation function is used to calculate the model's performance metrics on a test set, thereby deriving a comprehensive evaluation of the model. Both the evaluation function and the construction of the evaluation model are used to assess the reliability of the network optimization model. First, the network optimization model is trained using a training set, and its parameters and hyperparameters are adjusted. Then, the trained network optimization model is evaluated using a test set, and the evaluation function is calculated to obtain the network optimization model's performance metrics on unseen data. The purpose of model evaluation is to understand the generalization ability of the network optimization model, that is, its predictive ability on unknown data.

[0045] Step S300: Construct a network optimization model, use the evaluation model to evaluate the output information of the network optimization model, and determine whether the network optimization model has reached the preset reliability.

[0046] Furthermore, the process of constructing the network optimization model includes:

[0047] Step S310: Establish a finite element model of the part structure, and use the finite element model of the part structure to perform a preset number of iterations on the part to be optimized, and obtain the element update status and performance data of the part to be optimized during the iteration process;

[0048] Step S320: Construct a training dataset based on the unit update status and the performance data;

[0049] Step S330: Train the long short-term memory recurrent neural network based on the training dataset to obtain the trained network optimization model, and use the trained network optimization model to optimize the part to be optimized to obtain the output information.

[0050] Specifically, the network optimization model is a neural network-based optimization model. In this embodiment, preferably, an LSTM (Long Short-Term Memory) recurrent neural network optimization model is used to optimize the part structure. Before applying the network optimization model, the training and evaluation of the long short-term memory recurrent neural network model are first completed. First, a classic complex finite element model is established to optimize the part structure, and a preset number of iterations are performed using the classic topology optimization method to obtain the unit update status and performance data of the part to be optimized during multiple iterations. Among them, the unit update status includes update actions and new structures; the performance data includes parameters such as maximum equivalent stress, maximum displacement, and volume fraction V. After obtaining the above data, each data item is labeled, including labeling the update action as at, the new structure as ot, and the performance data of the new structure as F(X), thereby obtaining the training dataset {ot, at, F(X)}; this dataset is used as the training data for the long short-term memory recurrent neural network to train the model and obtain the final network optimization model.

[0051] Preferably, in this embodiment, when training the network optimization model, based on the aforementioned training dataset, a two-layer ANN (Artificial Neural Network) and backpropagation algorithm are used to train the Long Short-Term Memory Recurrent Neural Network to obtain the network optimization model. The first and second hidden layers of the network optimization model each have 25 nodes, and the ReLU activation function, Mean Absolute Error (MAE), and RMSPropOptimizer are used for optimization, respectively. The training parameters are set to a learning rate of 0.001 and a sample size of 1000. The hyperparameter values ​​of the network optimization model are shown in Table 1 below:

[0052] Table 1

[0053]

[0054] Among them, ReLU (Rectified Linear Unit) activation function is a commonly used non-linear function, often used in the hidden and output layers of neural networks; Mean Absolute Error (MAE) is a commonly used loss function in regression tasks, used to measure the average absolute difference between predicted and true values; RMSPropOptimizer is an optimizer in TensorFlow used to train neural network models. It is an implementation of the RMSProp (Root Mean Square Propagation) optimization algorithm. The RMSProp optimization algorithm aims to solve some problems in the gradient descent algorithm, especially the choice of learning rate and the large differences in convergence speed among different features. In the gradient descent algorithm, the learning rate is an important factor controlling the magnitude of parameter updates. However, using a fixed learning rate may lead to slow convergence or oscillations. The RMSProp optimization algorithm solves this problem by adaptively adjusting the learning rate. It uses an exponentially weighted average to calculate the moving average of the squared gradient of each parameter and uses it to update the parameters. Specifically, RMSProp uses a decay coefficient to control the weight of past squared gradients; a larger decay coefficient reduces the influence of past gradients, thus making the update smoother.

[0055] After obtaining the trained network optimization model, the network optimization model is used to optimize the part structure and output the part structure. The evaluation model is used to evaluate the output information of the network optimization model. By judging whether the output information of the network optimization model meets the constraints in the evaluation function, it is determined whether the network optimization model has reached the preset reliability.

[0056] Furthermore, in practical implementation, in addition to taking whether the evaluation function is satisfied as an influencing factor for evaluating the reliability of the network optimization model, factors such as the proportion of reliable structures that satisfy the constraints in the output information can also be added as influencing factors for evaluating the reliability of the network optimization model.

[0057] Step S400: If the network optimization model reaches the preset reliability, the network optimization model is used to optimize the structure of the part to be optimized to obtain an optimized structure.

[0058] Specifically, if the network optimization model reaches the preset reliability, the structure of the part to be optimized is optimized using the network optimization model. The process of structural optimization by the network optimization model includes: obtaining the updated model feature state (s), memorizing the feature state (h), generating an update scheme (a), and updating the memorized feature state (h); inputting the above parameters into the initial network optimization model, and after the network optimization model's rapid calculation, obtaining the performance data of the update scheme a and the new structure, and using the updated feature state as the next input, after the LSTM network optimization model's update iteration, obtaining the new structure that finally satisfies the loss function LOSS = Gt and outputting it, the output new structure is the optimized structure. By utilizing the sensitivity of the long short-term memory network optimization model to long-term and short-term memory, an output structure that can understand long-term and short-term memory is obtained, ultimately completing the update of the additive manufacturing part structure.

[0059] Based on the same inventive concept, this invention also provides an additive manufacturing part structure topology optimization system, as described in the following embodiments. Since the principle of solving the problem using an additive manufacturing part structure topology optimization system is similar to that of an additive manufacturing part structure topology optimization method, the implementation of an additive manufacturing part structure topology optimization system can refer to the implementation of an additive manufacturing part structure topology optimization method, and repeated details will not be elaborated further. As used below, the terms "unit" or "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the apparatus described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0060] Figure 3 This is a structural block diagram of an additive manufacturing part structure topology optimization system 300 according to an embodiment of the present invention, such as... Figure 3 As shown, it includes: an evaluation function acquisition module 310, used to define weight coefficients and constraints for different working conditions based on the working condition information of the part to be optimized, and to determine an evaluation function based on the weight coefficients and constraints; an evaluation model construction module 320, used to build an evaluation model based on the evaluation function; an optimization model construction module 330, used to construct a network optimization model, use the evaluation model to evaluate the output information of the network optimization model, and determine whether the network optimization model has reached a preset reliability; and a part structure optimization module 340, used to optimize the structure of the part to be optimized using the network optimization model if the network optimization model has reached the preset reliability, and to obtain an optimized structure.

[0061] Furthermore, the optimization model construction module 330 is also used to establish a finite element model of the part structure, use the finite element model of the part structure to perform a preset number of iterations on the part to be optimized, and obtain the unit update status and performance data of the part to be optimized during the iteration process; construct a training dataset based on the unit update status and the performance data; train a long short-term memory recurrent neural network based on the training dataset to obtain the network optimization model, and use the network optimization model to optimize the part to be optimized to obtain the output information.

[0062] In this embodiment, a computer device is provided, such as... Figure 4 As shown, it includes a memory 401, a processor 402, and a computer program stored in the memory 401 and executable on the processor 402. When the processor 402 executes the computer program, it implements any of the above-described additive manufacturing part structure topology optimization methods.

[0063] Specifically, the computer device can be a computer terminal, a server, or a similar computing device.

[0064] In this embodiment, a computer-readable storage medium is provided, which stores a computer program that executes any of the above-described additive manufacturing part structure topology optimization methods.

[0065] Specifically, computer-readable storage media, including both permanent and non-permanent, removable and non-removable media, can store information using any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer-readable storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable storage media does not include transient media, such as modulated data signals and carrier waves.

[0066] The embodiments of the present invention achieve the following technical effects:

[0067] 1. This invention constructs a new evaluation function that can be used in topology optimization applications of complex parts under multiple working conditions, and the results are more meaningful for engineering than those of traditional single working conditions.

[0068] 2. This invention proposes an LSTM network optimization model, which tracks the topology optimization process by using a constructed proxy model based on data obtained from a large number of classical topology optimization methods, resulting in more reliable results.

[0069] 3. Compared with the classic topology optimization method, the present invention has the advantage of fast calculation speed, especially for large and complex parts, avoiding complex finite element calculations and saving costs.

[0070] Obviously, those skilled in the art should understand that the modules or steps of the above-described embodiments of the present invention can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those presented here, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the embodiments of the present invention are not limited to any particular hardware and software combination.

[0071] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations of the embodiments of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method of topology optimization of an additively manufactured part structure, characterized in that, The application relates to a method for optimizing a part structure, comprising the following steps: According to the working condition information of the part to be optimized, the weight coefficients and constraint conditions under different working conditions are defined, and an evaluation function is determined according to the weight coefficients and the constraint conditions; An evaluation model is established according to the evaluation function G(X,A,ρ)=V(X)+ρ×AR(X); AR(X)=∑a i ×r i , wherein V(X) is a volume fraction, R(X) is a constraint condition, ρ is a punishment degree, AR(X) is a constraint condition evaluation item under multiple working conditions, A is a weight coefficient corresponding to different working conditions, a i is a weight coefficient under the i th working condition, and r i is a constraint condition under the i th working condition. A network optimization model is constructed, the output information of the network optimization model is evaluated by using the evaluation model, and it is judged whether the network optimization model reaches a preset reliability, wherein a part structure finite element model is established, the part to be optimized is iterated for a preset number of times by using the part structure finite element model, the unit updating condition and performance data of the part to be optimized in the iteration process are obtained, a training data set is constructed according to the unit updating condition and the performance data, a long short-term memory recurrent neural network is trained based on the training data set, the network optimization model is obtained, the part to be optimized is optimized by using the network optimization model, and the output information is obtained; If the network optimization model reaches the preset reliability, the structure of the part to be optimized is optimized by using the network optimization model, and an optimized structure is obtained, wherein in the process of training the long short-term memory recurrent neural network based on the training data set, the long short-term memory recurrent neural network is trained based on a back propagation algorithm, and a network optimization model is obtained, wherein the long short-term memory recurrent neural network comprises a first hidden layer and a second hidden layer; the node numbers of the first hidden layer and the second hidden layer are both 25, and the network optimization model is optimized by using a ReLU activation function, a mean absolute error (MAE) and an RMSPropOptimizer, respectively.

2. A method of topology optimization of a structure of an additively manufactured part according to claim 1, characterized in that, When the long short-term memory recurrent neural network is trained based on the training data set, the training parameter is set as a learning rate of 0.001 and a sample number of 1000.

3. A topology optimization system for additive manufacturing parts, characterized in that, The application relates to a method for optimizing a part structure, comprising the following steps: An evaluation function acquisition module is used for defining the weight coefficients and constraint conditions of different working conditions according to the working condition information of the part to be optimized, and an evaluation function is determined according to the weight coefficients and the constraint conditions; The evaluation model construction module is configured to establish an evaluation model according to the evaluation function G(X, A, p) = V(X) + p x AR(X); AR(X) = åa i r i wherein V(X) is a volume fraction, R(X) is a constraint condition, p is a punishment degree, AR(X) is a constraint condition evaluation item under multiple working conditions, A is a weight coefficient corresponding to different working conditions, a i i i is a weight coefficient under an i th working condition, and r i is a constraint condition under the i th working condition. An optimization model construction module is used for constructing a network optimization model, evaluating the output information of the network optimization model by using the evaluation model, and judging whether the network optimization model reaches a preset reliability, wherein a part structure finite element model is established, the part to be optimized is iterated for a preset number of times by using the part structure finite element model, the unit updating condition and performance data of the part to be optimized in the iteration process are obtained, a training data set is constructed according to the unit updating condition and the performance data, a long short-term memory recurrent neural network is trained based on the training data set, the network optimization model is obtained, the part to be optimized is optimized by using the network optimization model, and the output information is obtained; The part structure optimization module is configured to, if the network optimization model reaches the preset reliability, optimize the structure of the part to be optimized using the network optimization model to obtain an optimized structure. In the process of training the long short-term memory recurrent neural network based on the training data set, the long short-term memory recurrent neural network is trained based on a back propagation algorithm to obtain a network optimization model. The long short-term memory recurrent neural network includes a first hidden layer and a second hidden layer. The number of nodes of the first hidden layer and the second hidden layer is 25. The network optimization model is optimized using a ReLU activation function, a mean absolute error (MAE), and an RMSPropOptimizer, respectively.

4. The system of claim 3, wherein, The optimization model construction module is further configured to establish a part structure finite element model, perform a preset number of iterations on the part to be optimized using the part structure finite element model, and obtain an element update condition and performance data of the part to be optimized in the iteration process. The training data set is constructed based on the element update condition and the performance data. The long short-term memory recurrent neural network is trained based on the training data set to obtain the network optimization model. The part to be optimized is optimized using the network optimization model to obtain the output information.

5. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the additive manufacturing part structure topology optimization method of any one of claims 1 to 2.

6. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program for executing the additive manufacturing part structure topology optimization method of any one of claims 1 to 2.

Images