A Predicting Initial Layout of Components Method Using Machine Learning

In the structural topology optimization approaches, the Moving Morphable Components (MMC) is a new method to obtain the optimized structural topologies by optimizing shapes, sizes, and locations of components. However, the initial layout of components has a strong influence on the rate of convergence. In this paper, a predicting the initial layout of components method using machine learning is developed. In this method, the training set is generated under the MMC framework and supported vector regression (SVR) is employed to establish the mapping between the design parameters characterizing the initial layout and the number of iterations. How to combine machine learning (ML) with the MMC to predict the tilt angle initial layout of components that satisfy a given number of iterations is discussed. Finally, the cantilever beam example is sued to demonstrate the effectiveness of the proposed method.