Abstract: Monitoring composite fatigue damage is of great significance to ensure the safety of composite structures, for this reason, a composite fatigue damage monitoring method based on the combination of ultrasonic guided wave and machine learning model is proposed, aiming to improve the accuracy of damage identification. The method firstly screens out the ultrasonic guided wave signal features that are greatly affected by fatigue damage, then carries out covariance analysis to reduce the redundancy of the feature matrix, and preprocesses it to form a sample library. Subsequently, the signal feature matrix is used as the input, and the damage features are used as the output to construct a convolutional neural network(CNN) damage prediction model. This model is then combined with the bayesian optimization(BO) to adjust the hyperparameters of the model. Ultimately, the sample library is randomly divided into training set, validation set and test set according to the ratio of 8∶1∶1, using the validation set for hyper-parameter selection and the test set for model evaluation. The experimental results show that the constructed BO-CNN damage prediction model has better damage diagnosis capability relative to the CNN damage prediction model, the quantization accuracy is improved from 0.94 to 0.98, and the damage regression prediction task shows higher reliability.

Key words: structural health monitoring, fatigue damage, machine learning, convolutional neural network, Bayesian optimization