This project covers application of AI in structural analysis. Current Market of Artificial Intelligence is $ 22.6 Billion. Finite Element Analysis is used to perform structural analysis. Idea is to use the data set of structural analysis to predict the changes in system by Machine Learning method.
Proposal
The proposal is to make the datset of structural analysis displacement results and to machine learn the patterns to predict for other structures. Eventually with same approach by large data set FEA analysis can be performed more quickly in the future with AI.
Work Flow – Test 1
Made a script for parametric model of canopy. Then defined supports, load conditions and cross section information to send for FEA anlalysis with Karamaba. The best support connection position is unknown in the start, which is derived from evolutionary algorithm.
Process – Param
Made script for parametric model of canopy. Then defined supports, load conditions and cross section information to send for FEA anlalysis with Karamaba. The best support connection position is unknown in the start, which is derived from evolutionary algorithm.
FEA Analysis – Line To Beam Method
</p>
Work Flow – Test 2
Made a script for parametric model of canopy. Then defined supports, load conditions and cross section information to send for FEA anlalysis with Karamba. The best support connection position is unknown in the start, which is derived from evolutionary algorithm. Filtering Fittest Values of different canopy shapes for Neural Network learning.
Fittest Cases – FEA Analysis & Evolutionary Solver
Demo for different canopy design shapes.
Work Flow – Test 3
In this test the data set which is made from the fittest values of displacement is divided into Input and Output values for the Neural Network to train this data set.
Fittest Cases – FEA Analysis & Evolutionary Solver
Demo for different canopy design shapes.
</p>
Neural Network– Training
99 percent of data was trained from the data set values entered into neural network for training.
</p>
Neural Network– Custom Canopy Prediction
Predicting displacement of structure with trained data from Neural Network.
Machine Learning For Structure Patterns is a project of IAAC, Institute for Advanced Architecture of Catalonia developed in the Master in Robotics and Advanced Construction 2019/2020 by: Students: Mansoor Awais, Maria Isabel , Amit Patter Faculty: Mateusz Zwierzycki, Soroush Garivani