CN113343580A - Real-time topology optimization generation design method based on artificial intelligence technology - Google Patents
Real-time topology optimization generation design method based on artificial intelligence technology Download PDFInfo
- Publication number
- CN113343580A CN113343580A CN202110704660.8A CN202110704660A CN113343580A CN 113343580 A CN113343580 A CN 113343580A CN 202110704660 A CN202110704660 A CN 202110704660A CN 113343580 A CN113343580 A CN 113343580A
- Authority
- CN
- China
- Prior art keywords
- real
- topology optimization
- neural network
- result
- network model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000005457 optimization Methods 0.000 title claims abstract description 62
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000013461 design Methods 0.000 title claims abstract description 29
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 17
- 238000005516 engineering process Methods 0.000 title claims abstract description 15
- 238000003062 neural network model Methods 0.000 claims abstract description 25
- 238000012549 training Methods 0.000 claims abstract description 21
- 238000006073 displacement reaction Methods 0.000 claims abstract description 6
- 238000013528 artificial neural network Methods 0.000 claims description 8
- 238000013527 convolutional neural network Methods 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 6
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 5
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 238000012217 deletion Methods 0.000 claims description 3
- 230000037430 deletion Effects 0.000 claims description 3
- 230000003042 antagnostic effect Effects 0.000 claims description 2
- 230000008569 process Effects 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 11
- 238000010586 diagram Methods 0.000 description 3
- 238000012938 design process Methods 0.000 description 2
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/08—Probabilistic or stochastic CAD
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/10—Additive manufacturing, e.g. 3D printing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a real-time topological optimization generation design method based on artificial intelligence technology, which comprises the steps of (1) carrying out graphic coding on boundary conditions such as concentrated load, distributed load, displacement boundary conditions and the like; (2) generating a real topological optimization structure corresponding to the boundary condition selected in the step (1) by using an SIMP topological optimization method, and coding the real topological optimization structure; (3) storing the input condition labels x in the step (1) and the output result labels in the step (2) as training data sets after one-to-one correspondence, and providing the training data sets for subsequent neural network model training; (4) selecting a proper neural network model, taking the condition label x as input, taking the real topology optimization result y as an expected result, and training the neural network model; (5) and (3) coding the design conditions required to be applied in the actual design according to the coding format in the step (1), and inputting the coding conditions into the trained neural network model to generate a topology optimization result in real time. The method can generate a high-precision topological optimization result which is very similar to a result of a criterion method, can realize real-time generation, and greatly shortens the calculation time required by topological optimization.
Description
Technical Field
The invention relates to the technical field of topology optimization, in particular to a real-time topology optimization generation design method based on an artificial intelligence technology.
Background
At present, with the rapid development of equipment lightweight and additive manufacturing industry, the topology optimization is more and more valued by people. The purpose of topology optimization of structures is to find the most suitable material distribution within a given design domain, so that the optimized structure has some specific properties. However, the conventional topology optimization method has an inherent defect: "dimensionally cursing", i.e. when the design parameter variables and iteration steps increase suddenly, the time and memory required by the computer will increase exponentially, and all topology optimization algorithms face a difficult problem, i.e. how to effectively improve the computational efficiency.
In recent years, the artificial intelligence technology has been developed rapidly and is widely applied in different fields such as image recognition, natural language processing and the like. Artificial intelligence techniques are powerful in that they can learn certain features from data using neural network models, and trained models can make classification decisions or implement some function fits when faced with similar data again. The topology optimization and the artificial intelligence technology are combined, and the time consumed by the topology optimization can be greatly reduced by training a neural network model suitable for a topology optimization task in advance, so that the generation design of the real-time topology optimization can be realized.
Disclosure of Invention
Aiming at the problem of long time consumption of topology optimization design calculation, the invention aims to provide a real-time topology optimization generation design method based on an artificial intelligence technology.
The technical scheme adopted by the invention is as follows:
the invention provides a real-time topology optimization generation design method based on an artificial intelligence technology, which specifically comprises the following steps:
(1) carrying out graphic coding on boundary conditions such as concentrated load, distributed load, displacement boundary conditions and the like; regarding each condition as corresponding to a one-dimensional channel, connecting the one-dimensional channels end to serve as a model input vector, namely inputting a condition label x;
(2) generating a real topology optimization result corresponding to the boundary condition selected in the step (1) by using an SIMP topology optimization method, and carrying out graphic coding on the generated real topology optimization result to be used as an output result label y;
(3) storing the input condition labels x in the step (1) and the output result labels y in the step (2) as training data sets after one-to-one correspondence, and providing the training data sets for subsequent neural network model training;
(4) selecting a proper neural network model, taking the condition label x as input, taking the real topology optimization result y as an expected result, and taking the numerical difference between the real topology optimization result and the result generated by the neural network as a loss function in the training process; adjusting parameters of a neural network model through an optimizer, so that a loss function is reduced, and the output result of the neural network is close to the real topology optimization result y as much as possible;
(5) and (3) coding the design conditions required to be applied in the actual design according to the coding format in the step (1), and inputting the coding conditions into the trained neural network model to generate a topology optimization result in real time.
Further, the neural network model selects a convolutional neural network CNN model or a conditional generation antagonistic neural network CGAN model.
Furthermore, the neural network model is composed of a full connection layer, a random deletion layer, an anti-convolution layer, a convolution layer and an activation function layer.
Further, the optimizer selects a stochastic gradient descent algorithm SGD or an adaptive moment estimation algorithm Adam.
Further, the loss function is selected from a Mean Square Error (MSE) or a cross entropy loss function.
Further, the method is applicable to two-dimensional structures or three-dimensional structures.
Compared with the prior art, the invention has the following beneficial effects:
the invention combines the neural network model in the field of artificial intelligence with the topology optimization design, provides a method for eliminating pain points with long calculation time consumption in the topology optimization design process by pre-training the neural network model, can be used for quickly generating the topology optimization structure in real time, and greatly improves the efficiency of the actual design process.
Drawings
FIG. 1 is a flow chart of the technical solution of the present invention;
FIG. 2 is a schematic diagram of the model of the present invention;
FIG. 3 is a schematic diagram of a generator according to the present invention;
FIG. 4 is a schematic structural diagram of an arbiter according to the present invention;
FIG. 5 is a comparison of the topology optimization results generated by the present invention and the topology optimization results generated by the SIMP method.
Detailed Description
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
The real-time topology optimization generation design method based on the artificial intelligence technology can select a proper neural network model for training aiming at different topology optimization design requirements, so that real-time topology optimization generation design is realized. In the following, a model for generating and designing real-time topology optimization is trained based on a conditional generation countermeasure neural network CGAN by taking two-dimensional structure topology optimization as an example.
As shown in fig. 1 and fig. 2, the method comprises the following specific steps:
(1) carrying out graphic coding on boundary conditions such as concentrated load, distributed load, displacement boundary conditions and the like; considering each condition as corresponding to a one-dimensional channel, the channel comprising: 2 load boundary condition channels and 2 displacement boundary condition channels, wherein the one-dimensional channels are connected end to end and then used as model input vectors, namely conditional feature labels x; if the size of the design domain is 80 × 80, only considering that the number of the pixel points on the boundary is 79 × 4, namely 316, and because 2 load boundary condition channels and 2 displacement boundary conditions are input and 4 boundary conditions are total, the size of the corresponding conditional feature label x is 316 × 4, namely a one-dimensional vector with the length of 1264;
(2) generating a real topology optimization result corresponding to the boundary condition selected in the step (1) by using a SIMP topology optimization method, namely, taking an array with the size of 80 multiplied by 80 as a result label y;
(3) storing the condition label x in the step (1) and the result label y in the step (2) as a training data set after corresponding to each other for use in subsequent training of the neural network model;
(4) selecting a proper neural network model, wherein the neural network model can select a convolutional neural network CNN or a condition to generate a model which is suitable for design requirements, such as an anti-neural network CGAN and the like; the embodiment selects a generation confrontation neural network model CGAN which consists of a generator and a discriminator;
wherein, the generator structure: the first part of network is a full connection layer, and the parameters are 5 multiplied by 8 multiplied by D; the second part of the network is a Dropout random node deletion layer, and the parameters are [5, 5, 8 × D ]; the third part of the network is 4 deconvolution modules, and each deconvolution module comprises a Deconv2D deconvolution layer and a Relu activation function layer; the parameter settings of each convolutional layer are [10, 10, 4 × D ], [20, 20, 2 × D ], [40, 40, D ], [80, 80, 1], respectively; the structure and parameters of the generator are shown in fig. 3; d is the weight of the channel number, and can be changed at will according to the size of the memory;
the structure of the discriminator: the first part of the network is 4 convolution modules, each convolution module comprises a Conv2D convolution layer and an LRelu activation function layer, the parameters of each convolution layer are [44, 44, D ], [22, 22, 2 xD ], [11, 11, 4 xD ], [6, 6, 8 xD ], and the parameter of the LRelu activation function layer is 0.2; the second partial network is a full connectivity layer with final input of 0/1; the structure and parameters of the generator are shown in fig. 4; d is the weight of the channel number, and can be changed at will according to the size of the memory;
inputting the conditional feature label x into a generator, and generating a false topology optimization structure by using the generator;
splicing the false topology optimization structure and the corresponding real topology optimization structure generated by the generator with the corresponding conditional feature label x respectively, and completing the spliced vectors by using 0 to reconstruct the vectors into a square gray-scale image with the size of [88, 88, 1 ];
inputting the reconstructed square gray scale image into a discriminator to judge authenticity;
the method comprises the steps that an optimizer is used for alternately training a generator and a discriminator, the optimizer can adopt an optimizer which is easy to converge such as a random gradient descent algorithm SGD or an adaptive moment estimation algorithm Adam, and the like, in the embodiment, the ADAM optimizer is used for alternately training the generator and the discriminator, wherein the discriminator trains the generator for 1 time every 5 times, and 100 iteration rounds are designed; wherein the discriminant loss function and the ADAM optimizer are expressed as follows:
where D (x) is the probability that x is considered to be true by the discriminator and λ is the gradient penalty coefficient.
Wherein Dw(Gθ(z)) is the probability that the data generated by the generator is considered to be true by the discriminator, α is the learning rate, β1Exponential decay Rate, beta, estimated for the first moment2The exponential decay rate estimated for the second moment.
(5) After the model training is finished, the boundary conditions required to be applied in the actual design can be coded and input into the generator according to the method shown in the step (1), and then the corresponding topology optimization structure can be generated in real time; the pair of topologically optimized structures actually generated using the present method versus those produced by SIMP is shown in fig. 5.
The generation countermeasure model combining the denoised CGAN and the WGAN-GP is used for generating a topology optimization structure in real time, removes random noise in the traditional CGAN, and reduces the number of model parameters by only using input as conditional characteristics; the Wasserstein distance is used as a loss function, so that the training difficulty of the model is effectively reduced.
The method provided by the invention is not only suitable for the two-dimensional structure topology optimization in the above embodiments, but also suitable for the three-dimensional structure topology optimization. The above-mentioned embodiments are merely illustrative of the preferred embodiments of the present invention, and do not limit the scope of the present invention, and various modifications and improvements of the technical solution of the present invention by those skilled in the art should fall within the protection scope defined by the claims of the present invention without departing from the spirit of the present invention.
Claims (6)
1. A real-time topology optimization generation design method based on artificial intelligence technology is characterized by comprising the following steps:
(1) carrying out graphic coding on boundary conditions such as concentrated load, distributed load, displacement boundary conditions and the like; regarding each condition as corresponding to a one-dimensional channel, connecting the one-dimensional channels end to serve as a model input vector, namely inputting a condition label x;
(2) generating a real topology optimization result corresponding to the boundary condition selected in the step (1) by using an SIMP topology optimization method, and carrying out graphic coding on the generated real topology optimization result to be used as an output result label y;
(3) storing the input condition labels x in the step (1) and the output result labels y in the step (2) as training data sets after one-to-one correspondence, and providing the training data sets for subsequent neural network model training;
(4) selecting a proper neural network model, taking the condition label x as input, taking the real topology optimization result y as an expected result, and taking the numerical difference between the real topology optimization result and the result generated by the neural network as a loss function in the training process; adjusting parameters of a neural network model through an optimizer, so that a loss function is reduced, and the output result of the neural network is close to the real topology optimization result y as much as possible;
(5) and (3) coding the design conditions required to be applied in the actual design according to the coding format in the step (1), and inputting the coding conditions into the trained neural network model to generate a topology optimization result in real time.
2. The real-time topology optimization generation design method based on artificial intelligence technology according to claim 1, characterized in that: the neural network model selects a convolutional neural network CNN model or a condition to generate an antagonistic neural network CGAN model.
3. The real-time topology optimization generation design method based on artificial intelligence technology according to claim 1 or 2, characterized in that: the neural network model is composed of a full connection layer, a random deletion layer, an anti-convolution layer, a convolution layer and an activation function layer.
4. The real-time topology optimization generation design method based on artificial intelligence technology according to claim 1, characterized in that: the optimizer selects a random gradient descent algorithm SGD or an adaptive moment estimation algorithm Adam.
5. The real-time topology optimization generation design method based on artificial intelligence technology according to claim 1, characterized in that: the loss function is selected from a Mean Square Error (MSE) or a cross entropy loss function.
6. The real-time topology optimization generation design method based on artificial intelligence technology according to claim 1, characterized in that: the method is applicable to two-dimensional structures or three-dimensional structures.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110704660.8A CN113343580A (en) | 2021-06-24 | 2021-06-24 | Real-time topology optimization generation design method based on artificial intelligence technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110704660.8A CN113343580A (en) | 2021-06-24 | 2021-06-24 | Real-time topology optimization generation design method based on artificial intelligence technology |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113343580A true CN113343580A (en) | 2021-09-03 |
Family
ID=77478437
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110704660.8A Pending CN113343580A (en) | 2021-06-24 | 2021-06-24 | Real-time topology optimization generation design method based on artificial intelligence technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113343580A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108280305A (en) * | 2018-01-30 | 2018-07-13 | 西安交通大学 | Radiating element cooling duct rapid topology optimum design method based on deep learning |
| CN109168003A (en) * | 2018-09-04 | 2019-01-08 | 中国科学院计算技术研究所 | A method of generating the neural network model for being used for video estimation |
| CN109783910A (en) * | 2018-12-29 | 2019-05-21 | 西安交通大学 | It is a kind of to utilize the optimum structure design method for generating confrontation network acceleration |
| CN111723420A (en) * | 2020-05-20 | 2020-09-29 | 同济大学 | Structural topology optimization method based on deep learning |
| CN111898730A (en) * | 2020-06-17 | 2020-11-06 | 西安交通大学 | Structure optimization design method for accelerating by using graph convolution neural network structure |
-
2021
- 2021-06-24 CN CN202110704660.8A patent/CN113343580A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108280305A (en) * | 2018-01-30 | 2018-07-13 | 西安交通大学 | Radiating element cooling duct rapid topology optimum design method based on deep learning |
| CN109168003A (en) * | 2018-09-04 | 2019-01-08 | 中国科学院计算技术研究所 | A method of generating the neural network model for being used for video estimation |
| CN109783910A (en) * | 2018-12-29 | 2019-05-21 | 西安交通大学 | It is a kind of to utilize the optimum structure design method for generating confrontation network acceleration |
| CN111723420A (en) * | 2020-05-20 | 2020-09-29 | 同济大学 | Structural topology optimization method based on deep learning |
| CN111898730A (en) * | 2020-06-17 | 2020-11-06 | 西安交通大学 | Structure optimization design method for accelerating by using graph convolution neural network structure |
Non-Patent Citations (2)
| Title |
|---|
| 刘笑辰: "基于深度学习的实时拓扑优化设计方法", 《CNKI优秀硕士学位论文全文库》, 15 February 2020 (2020-02-15), pages 1 - 70 * |
| 刘笑辰: "基于深度学习的实时拓扑优化设计方法", 《硕士学位论文》, 15 February 2020 (2020-02-15), pages 1 - 70 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110059878B (en) | Photovoltaic power generation power prediction model based on CNN LSTM and construction method thereof | |
| CN107516129B (en) | Dimension self-adaptive Tucker decomposition-based deep network compression method | |
| CN113905391B (en) | Integrated learning network traffic prediction method, system, equipment, terminal and medium | |
| CN113065649B (en) | Complex network topology graph representation learning method, prediction method and server | |
| Shao et al. | Branchy-GNN: A device-edge co-inference framework for efficient point cloud processing | |
| CN114662788B (en) | Seawater quality three-dimensional time-space sequence multi-parameter accurate prediction method and system | |
| CN112215353A (en) | Channel pruning method based on variational structure optimization network | |
| CN115471423A (en) | Point cloud denoising method based on generation countermeasure network and self-attention mechanism | |
| CN116484740A (en) | Line parameter identification method based on space topology characteristics of excavated power grid | |
| CN115860215A (en) | Photovoltaic and wind power generation power prediction method and system | |
| CN112329918A (en) | Anti-regularization network embedding method based on attention mechanism | |
| CN117036901A (en) | Small sample fine adjustment method based on visual self-attention model | |
| CN113343580A (en) | Real-time topology optimization generation design method based on artificial intelligence technology | |
| Adigun et al. | Training generative adversarial networks with bidirectional backpropagation | |
| CN115906959A (en) | Parameter training method of neural network model based on DE-BP algorithm | |
| CN115270917A (en) | Two-stage processing multi-mode garment image generation method | |
| Sarkar et al. | An incremental pruning strategy for fast training of CNN models | |
| CN113807497A (en) | Non-paired image translation method for enhancing texture details | |
| Liu et al. | A deep neural network pruning method based on gradient L1-norm | |
| Shen et al. | Learning to prune in training via dynamic channel propagation | |
| CN113111308A (en) | Symbolic regression method and system based on data-driven genetic programming algorithm | |
| Li et al. | Fast scenario reduction for power systems by deep learning | |
| CN114092579B (en) | Point cloud compression method based on implicit neural network | |
| CN112116062A (en) | Multilayer perceptron nonlinear compression method based on tensor string decomposition | |
| Mungai et al. | A study on merging mechanisms of simple hopfield network models for building associative memory |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |