Prediction of mechanical property of open-hole composite laminates using generalized regression neural network method

Abstract

Mechanical connection is a common method used for joining composite materials, but it is bound to open holes in the composite material structure. These open holes may cause stress concentration at the hole edge, impacting the overall mechanical properties of the component. In this paper, a machine learning-based method for predicting the mechanical properties of open-hole composite laminates is proposed based on generalized regression neural network. In detail, by using the Hashin failure criterion, the finite element models of composite laminates with single holes of different diameters have been established. Their load–displacement curves, maximum failure stresses and maximum failure strains are calculated numerically. Then, the different hole diameters and corresponding load–displacements can be used as the input and output variables of the generalized regression neural network to train the neural network model. Based on the optimal generalized regression neural network model, the mechanical properties of the composite laminates with a certain single hole diameter can be predicted. Compared with the uniaxial tensile experiment of open-hole composite laminates, the effectiveness of this machine learning method is verified. Furthermore, the changes in mechanical properties of double-hole composite laminates under different hole diameters and positions are analyzed. This study holds significant practical implications for enhancing the understanding of the mechanical properties of composite materials and the influence of defects on their performance.