Fig.
Star setting: enter the official account homepage, click the “top” corner, click “set star”, and the official account will appear yellow Pentagram (Android The method is introduced in a 6-minute small video (the same as the operation of IOS users).
After the calculation, the overall indexes such as the amount of structural materials, the shear weight ratio and the maximum inter story displacement angle can be obtained.
Invention Name: shear wall structure layout method and device based on Countermeasure generation network inventor: Lu Xinzheng; Liao Wenjie; Zheng zhe patent No.: zl202010446468.91 patent overview intelligent structural design is an important content of intelligent construction.
It can be seen that the difference between the structural design completed by this method and that designed by engineers is very small.
In this study, a large number of architectural design drawings and corresponding structural drawings are input into the neural network.
Figure 11 shear wall structure scheme designed by artificial intelligence imports the generated structure scheme (Figure 11) into the parametric modeling tool swallow of the engineering digital technology center of Central South Architectural Design Institute (see: building black technology | building structure auxiliary design software), which can automatically generate the required structure model by adjusting the parameters such as wall thickness, floor height and number of floors.
The flat drawing of the house type is processed with reference to the required pre-processing method of architectural design drawing to obtain the pixel diagram of architectural design.
The preliminary feature extraction process of typical design drawings is shown in Figure 6.
When the reader opens this article, the modeling of boundary conditions and loads has been automatically completed, the structural calculation has been completed and the calculation sheet has been generated (Figure 13), which takes about 3 minutes.
One neural network is called “generator”, A new image can be generated according to some input information (the input information in this patent is architectural drawing (house type layout plan), and the generated is structural scheme drawing).
In addition, in the last week, many experts from design institutes provided us with their shear wall residential building scheme.
Is it possible for artificial intelligence to learn the design experience of engineers by learning a lot of existing design data, so as to enable computers to do structural scheme design? Based on the confrontation generation network model, the invention proposes a shear wall structure layout method based on the confrontation generation network, as shown in Fig.
Another neural network is called “discriminator”, which is responsible for judging whether the image generated by the generator meets the requirements.
1.
1 generative design method of shear wall structure based on Countermeasure generation network Fig.
4 typical case application using this method, we have verified the application of several architectural designs, and some typical results are shown in Figure 8.
3 multiple case application effects of intelligent design method 2 what is countermeasure generation network? The core of this method is “countermeasure generation network”, and its core is two neural networks.
For example, extract three key elements in the architectural design drawings: the location of shear wall, indoor door and window openings and outdoor door openings; The key elements are filled with different colors, in which gray represents the location of the shear wall, green represents the indoor door and window openings, and blue represents the outdoor door openings.
Fig.
We generated the corresponding structural scheme with artificial intelligence.
If the discriminator cannot judge whether the drawing is designed by artificial intelligence or engineers, the discriminator will be judged as “failed” and go back to learn and improve.
Input the architectural design pixel diagram into our artificial intelligence structural scheme design tool structgan for structural scheme design.
If the discriminator finds that the structural scheme is designed by artificial intelligence rather than by engineers, the generator is judged as “failed” and the generator continues to learn and improve.
The small video of software operation demonstration takes only 5 minutes from the design of structural scheme to the completion of calculation sheet, and the video is not accelerated: the first step is to clarify the design task.
Figure 13 automatic generation of calculation sheet through the above method, after the architect determines the architectural scheme, the structural scheme layout and main structural economic indicators can be obtained within 5 minutes, and the efficiency of the whole design, modeling and analysis process has been significantly improved.
On the other hand, the discriminator constantly compares the structural scheme generated by artificial intelligence with the scheme designed by engineers.
8 application results of some typical cases Fig.
The performance evaluation method proposed by the invention is used to evaluate the countermeasure generation network model, which can be applied after passing the evaluation.
At present, the structural scheme design mainly depends on the experience of designers and lacks intelligent design methods.
A series of case designs have been carried out with the invention, and the typical effects are shown in Fig.
4) Fig.
The discriminator and generator continue to fight and improve until they reach the Nash equilibrium, which is called “confrontation generation network”.
The structural model generated by the parametric modeling tool in Figure 12 further calls ETABS software to calculate the structural model in Figure 12.
Fig.
3.
Fig.
We also invited some experts to judge and score the results of artificial intelligence design and engineer design.
On the one hand, the generator constantly learns how to generate the structural scheme according to the input architectural drawings.
3.2 step S102: the features of the architectural design drawings are preliminarily extracted, the key elements in the architectural design drawings are extracted, and the key elements are filled with different colors to generate the image features to be input.
5 using the intelligent generative structural design method step 3.1 step S101 obtain the architectural design drawings to be processed, and obtain the CAD drawings of shear wall residential architectural design that need structural design.
9 expert evaluation results based on blind test questionnaire case demonstration of artificial intelligence design of shear wall structure has attracted the attention of many friends because of the structural scheme design method based on artificial intelligence.
Figure 10 plan of shear wall residence when the reader opens this article, the structural scheme design of the above shear wall residence has been completed by the artificial intelligence structural scheme design tool structgan (Figure 11), which takes about 1 minute.
7 pre training of countermeasure generation network model 3.4 step S103 shear wall structure design drawing generation based on the countermeasure generation network model trained in step s102.5, input the building feature map obtained in step S102 to directly obtain the structural design drawing (taking several minutes).
2 typical case application effect of intelligent design method Fig.
The architect has designed a 30 storey shear wall residence, and the flat layout of the house type is shown in Figure 10.
7) until the design effect is stable.
4 basic principle of countermeasure generation network 3 patent use steps the main use steps of the patent are shown in Fig.
When readers open this article and read here, the structural model of the whole shear wall residence has been established, which takes 1 minute.
The invention preliminarily realizes the goal of enabling artificial intelligence to learn engineers’ design experience and improve design efficiency.
Some peer experts contacted the research group to further understand the operation mode and effect of this method.
Because the official account official has adjusted the push rules, if you want to continue to receive the contents of the official account, please set the public building number of the building as the star sign, and more points of the “look at” and “praise” in the end of the article.
Therefore, Lu Xinzheng research group of Tsinghua University cooperated with the engineering digital technology center of Central South Architectural Design Institute to make a case of using artificial intelligence method to design shear wall structure for your reference.
The results show that AI design can almost confuse the false with the true (Fig.
2 and Fig.
“If you want to continue to receive the official account, please note that the” WeChat “is a public key.
Some typical cases are shown in the figure below to show the ability of artificial intelligence to generate shear wall structural scheme..
6 preliminary feature extraction of architectural design drawing 3.3 step s102.5 countermeasure generation network model training adopts the feature extraction method in step S102 to extract the features of the supporting drawings of the existing shear wall building structure design, and then construct the training and test data set; Then, the countermeasure generation network model is trained based on the data set (Fig.
5, and the use will be introduced in detail later.
9).
(Fig.
Fig.
