【目的】随着智能船舶的发展，传统裂纹损伤检测方法难以满足检测需求。本文提出了一种基于门控循环单元（gated recurrent unit, GRU）神经网络的箱型梁结构裂纹损伤实时检测方法。
【结果】与基于小波包变换的多层感知机神经网络（wavelet packet transform-based multi-layer perceptron, WPT-MLP）进行对比，本文提出的GRU模型在损伤位置和损伤长度的检测上都具有更高的检测精度，且对噪声的敏感程度更低，对损伤位置的近似预测也有较高的精度。
[Objectives] With the development of intelligent ships, traditional crack damage detection methods are difficult to meet the detection requirements. This paper proposes a real-time crack damage detection method for box-beam based on gated recurrent unit neural network(gated recurrent unit, GRU).
[Methods] Using the secondary development technology of Abaqus based on Python language, the box beam crack damage model is built, and its acceleration response under dynamic Gaussian white noise excitation is calculated. The data set is generated by expanding the original data using data cropping method, and the influence of noise is considered. A box-beam crack damage detection model based on GRU is established, and the acceleration response data set is directly used as input. The minimum loss function value is used as target to train the model.
[Results] The comparison with the multi-layer perceptron neural network based on wavelet packet transform (wavelet packet transform-based multi-layer perceptron, WPT-MLP) shows that the GRU model proposed in this paper has higher detection accuracy in damage location detection and damage length detection. The model is less sensitive to noise and has high accuracy in the approximate prediction.
[Conclusions] The applicability of GRU neural network in crack damage detection of box-beam containing multiple plates is proved.