【摘要】 为准确识别波浪砰击载荷，保障海洋平台作业的安全性，基于卷积神经网络（CNN）模型，建立立柱-甲板简化模型的砰击载荷识别方法，根据结构测点应变响应数据对砰击载荷进行识别。通过分析半潜式平台立柱-甲板处砰击载荷分布的特点，进行有限元数值计算，生成训练数据集和测试数据集，同时考虑数据噪声的影响。结果表明：CNN砰击载荷识别模型具有很高的识别精度和良好的抗噪声能力，在无噪声和有噪声情况下，整体精度分别为99.6%和91.9%；相比传统的反向传播神经网络（BPNN）模型，CNN模型的识别精度更高。更多还原

【Abstract】 In order to accurately identify the wave slamming load for the safety of offshore platforms, a wave slamming load identification method is proposed for the simplified column-deck model based on the convolution neural network(CNN), and the wave slamming load can be identified according to the strain response of the structure. By analyzing the load distribution characteristics of the semi-submersible platform column-deck area, the finite element numerical simulations are carried out to generate the training set and test set, considering the influence of noise. The results show that the overall accuracy of CNN wave slamming load identification model is 99.6% and 91.9% in the cases with noise and without noise, respectively, indicating the high accuracy and good anti-noise ability. Compared with the traditional back propagation neural network(BPNN) model, CNN method has higher load identification accuracy for both noiseless and noisy conditions.更多还原