Impact load identification of nonlinear structures using deep Recurrent Neural Network

In this paper, a novel impact load identification method of nonlinear structures by using deep Recurrent Neural Network (RNN) is proposed. The deep RNN model, mainly consisting of two Long Short-Term Memory (LSTM) layers and one bidirectional LSTM (BLSTM) layer, is trained through a large number of dynamic responses and impact loads to learn the complex inverse mapping between structural inputs and outputs. The effectiveness and practicability of the proposed method are verified by three nonlinear cases: damped Duffing oscillator, nonlinear three-degree-of-freedom system and nonlinear composite plate. The results show that the proposed method has the capability for identifying the complex impact load even when the impact location is unknown. Meanwhile, hyperparameters of the deep RNN model and placement scheme of sensors are not highly sensitive to the identification accuracy. (C) 2019 Elsevier Ltd. All rights reserved.