An Artificial Intelligence-Based Failure Prediction Model for Three-Dimensional Woven Composite of Aircraft Wings

Three-dimensional (3D) woven composites are used to fabricate aerospace structures due to their lightweight properties. Structures made of 3D woven composite have a risk of failure depending on the number of fill yarn layers, which is an important design parameter. It is difficult to predict failure since the 3D woven composites have complex patterns. This paper proposes a method to efficiently predict the failure of 3D woven composite aircraft wings using an artificial intelligence (AI) model. We calculated the mechanical properties of each part that makes up the 3D woven composite wing structures using analytical methods. The yarn pattern is defined using geometric parameters and functions to obtain the mechanical properties. The stiffness and Poisson’s ratio were computed from a weighted average model (WAM) that combines iso-strain and iso-stress assumptions. The strength was obtained by increasing the load based on the local and global failure criteria. The failure index dataset was built using numerical methods. The failure index data are divided into training, validation, and test. An artificial neural network (ANN) learns the training data by forward and backward propagation. Epochs repeat until the errors are minimized through the loss function. The accuracy of the ANN is determined by an f-1 score and an area under the curve (AUC). We compared the results with the accuracy of other machine learning methods. Predicting the failure of the aircraft wings using the ANN is more time efficient than using different failure predicting methods.