对海洋浮式平台系泊系统在复杂多变的作业环境受到的线性、非线性作用力进行了预测,预测过程中对长短周期记忆(LSTM)单模型预测网络隐藏层数、迭代次数和学习速率做了优化。提出了具有可变卷积和小波基激活函数的多层特征提取特性和变阈值残差收缩预测功能的混合预测模型,对平台运动响应多点系泊系统整体受力进行非线性映射,分析了多点系泊缆模型,得到了风浪流联合作用下的系泊缆张力值。使用LSTM单模型、Resnet-LSTM混合模型和改进混合模型对系泊力仿真数据集进行训练预测。结果显示:采用Resnet-LSTM混合模型预测准确度可达0.9974,使用可变卷积改进的Resnet-LSTM预测效果优于未改进模型,各项网络参数和预测指标得以优化。证明基于Resnet-LSTM的改进混合预测模型应用在多点系泊系统张力非线性时序特征预测应用方面具有提升网络性能的作用。
Abstract
The linear and nonlinear forces on an offshore floating platform mooring system in a complex and variable operating environment are predicted, and the number of hidden layers, iterations, and learning rate of the long-short cycle memory (LSTM) single-model prediction network are optimized during the prediction process. A hybrid prediction model with multilayer feature extraction characteristics of variable convolution and wavelet-based activation functions and variable threshold residual shrinkage prediction function is proposed to nonlinearly map the overall forces of the multi-point mooring system in response to the platform motion, analyze the model of multi-point mooring cables, and obtain the value of the mooring cable tension under the combined action of wind and wave currents. The mooring force simulation dataset is trained for prediction using LSTM single model, Resnet-LSTM hybrid model and improved hybrid model. The results show and the network parameters and prediction indicators are optimized, that the prediction accuracy can be as high as 0.9974 using the Resnet-LSTM hybrid model, and the improved Resnet-LSTM prediction using variable convolution is better than the unimproved model. It is demonstrated that the application of the improved hybrid prediction model based on Resnet-LSTM has the effect of improving the network performance in the application of nonlinear time-series feature prediction of tension in multipoint mooring systems.
关键词
力学计量 /
系泊力 /
多点系泊系统 /
改进Resnet-LSTM模型 /
张力预测
Key words
mechanica metrology;mooring lines tension /
multi-points mooring system /
improved Resnet-LSTM model /
tension prediction
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1]黄晨亮, 郭力群, 吕阳阳, 等. 基于融合RBF-PSO-AE算法的混凝土力学性能预测 [J]. 计量学报, 2022, 43(11): 1464-1469.
HUANG C L, GUO L Q, L Y Y, et al. Prediction of concrete mechanical properties based on fusion RBF-PSO-AE algorithm [J]. Acta Metrologica Sinica, 2022, 43(11): 1464-1469.
[2]迟玉伦, 吴耀宇, 江 欢, 等. 基于模糊神经网络与主成分分析的磨削表面粗糙度在线预测 [J]. 计量学报, 2022, 43(11): 1389-1397.
CHI Y Y, WU Y Y, JIANG H, et al. Online prediction of grinding surface roughness based on fuzzy neural network and principal component analysis [J]. Acta Metrologica Sinica, 2022, 43(11): 1389-1397.
[3]金秀章, 李京. 基于互信息PSO-LSSVM的SO2浓度预测 [J]. 计量学报, 2021, 42(5): 675-680.
JIN X Z, LI J. Prediction of SO2 concentration based on mutual information PSO-LSSVM [J]. Acta Metrologica Sinica, 2021, 42(5): 675-680.
[4]卢锦玲, 郭鲁豫. 基于改进深度残差收缩网络的电力系统暂态稳定评估 [J]. 电工技术学报, 2021, 36(11): 2233-2244.
LU J L, GUO L Y. Power System Transient Stability assessment Based on improved Deep Residual Shrinkage Network [J]. Transactions of China Electrotechnical Society, 2021, 36(11): 2233-2244.
[5]何涛, 陈剑, 闻英友. 基于深度残差收缩网络的HEp-2图像识别 [J]. 计算机与现代化, 2021(1): 38-42.
HE T, CHEN J, WEN Y Y. HEp-2 Image Recognition Based on Deep Residual Shrinkage Network [J]. Jisuanji Yu Xiandaihua, 2021(1): 38-42.
[6]宋子豪, 程 伟, 彭岑昕, 等. 基于CWD和残差收缩网络的调制方式识别方法 [J]. 系统工程与电子技术, 2021, 43(11): 3371-3379.
SONG Z H, CHENG W, PENG Q X, et al. Modulation recognition based on CWD and residual shrinkage network, 2021, 43(11): 3371-3379.
[7]车畅畅, 王华伟, 倪晓梅, 等. 基于深度残差收缩网络的滚动轴承故障诊断 [J]. 北京航空航天大学学报, 2021, 47(7): 1339-1405.
CHE C C, WANG H W, NI X M, et al. Fault diagnosis of rolling bearing based on deep residual shrinkage network [J]. Journal of Beijing University of Aeronautics and Astronautics, 2021, 47(7): 1339-1405.
[8]张力, 常俊, 武浩, 等. 深度残差收缩网络下的定位与行为联合识别 [J]. 计算机工程与应用, 2021(6): 1-11.
ZHANG L, CHANG J, WU H, et al. Joint Location and Behavior Recognition in Deep Residual Shrinkage Network [J]. Computer Engineering and Applications, 2021(6): 1-11.
[9]CHUNG M, KIM S, LEE K. Detection of damaged mooring line based on deep neural networks [J]. Ocean Engineering, 2020,209:107522.
[10]MAO Y X, WANG T Q, DUAN M L. A DNN-based approach to predict dynamic mooring tensions for semi-submersible platform under a mooring line failure condition [J]. Ocean Engineering, 2022(9): 1-20.
[11]ALI M, PRASAD R, XIANG Y, et al. Advanced extreme learning machines vs. deep learning models for peak wave energy period forecasting: A case study in Queensland, Australia [J]. Renewable Energy, 2021,177: 1031-1044.
[12]XU S, JI C Y, SOARES C G. Short-term extreme mooring tension and uncertainty analysis by a modified ACER method with adaptive Markov Chain Monte Carlo simulations [J]. Ocean Engineering, 2021:109445.
[13]邓林青, 朱耀文, 王宏伟, 等. 基于NARX神经网络的FPSO系泊缆张力预报 [J]. 海洋工程装备与技术, 2020(4): 85-92.
DENG L Q, ZHU Y W, WANG H W, et al. NARX Neural Network Based FPSO Mooring Lines Tension Prediction [J]. Ocean Engineering Equipment Technology, 2020(4): 85-92.
[14]ZHANG D Q, WANG C, DU J F. A novel method for predicting dynamic response of deep-sea floating system based on artificial neural network [J]. Ship Building of China, 2021(3): 123-132.
[15]韩宇, 黄国良, 李鹏, 等. 基于LSTM的软刚臂单点系泊缆张力预测方法研究 [J]. 天津科技, 2020(5): 74-80.
HAN Y, HUANG G L, LI P, et al. Research on Prediction Method of Single-point Mooring Force of Soft Yoke Based on LSTM Method [J]. Tianjing Science & Technology, 2020(5): 74-80.
[16]YANG Y, PENG T, LIAO S J. Predicting future mooring line tension of floating structure by machine learning [J]. Ocean Engineering, 2022(9): 1-15.
[17]WANG Z M, QIAO D S, YAN J, et al. A new approach to predict dynamic mooring tension using LSTM neural network based on responses of floating structure [J]. Ocean Engineering, 2022: 110905.
基金
浙江省自然科学基金(LY19E090004)
PDF(648 KB)
