Wave Run-Up Prediction of Semi-Submersible Platforms Based on Long Short-Term Memory Network

doi:10.16183/j.cnki.jsjtu.2021.310

Abstract

Abstract:

Wave run-up and air-gap are key issues for the safety of semi-submersible platforms. Real-time wave run-up prediction is helpful to ensure the safety of offshore activities. Based on the long short-term memory (LSTM) network, the extreme short term online prediction method is developed for predicting the wave run-up of semi-submersible platforms using wave and motion sequences. With the help of large sets of data from the model test, the LSTM model is trained and tested. The study shows that when the forecast durations are 6 s and 12 s, the average accuracy of the prediction results are 92.90% and 84.09%, and the relative errors of the maximum wave run-up height are lower than or equal to 19.69% and 30.66%, respectively. In addition, the model has a stable and exact prediction of extreme values of wave run-up height when the forecast duration is within 6 s, which confirms its ability to provide valid technical support for the early warning of wave slamming and overtopping during the operation of offshore platforms.

Key words: long short-term memory (LSTM) network, wave run-up, online prediction, semi-submersible platform

CLC Number:

U674.38

LI Yan, XIAO Longfei, WEI Handi, KOU Yufeng. Wave Run-Up Prediction of Semi-Submersible Platforms Based on Long Short-Term Memory Network[J]. Journal of Shanghai Jiao Tong University, 2023, 57(2): 161-167.

Figures/Tables 12

Fig.1

Fig.2

Tab.1

Fig.3

Fig.4

Tab.2

Tab.3

Fig.5

Fig.6

Tab.4

Tab.5

Fig.7

References 11

[1]	单铁兵. 波浪爬升的机理性探索和半潜式平台气隙响应的关键特性研究[D]. 上海: 上海交通大学, 2013.
	SHAN Tiebing. Research on the mechanism of wave run-up and the key characteristics of air-gap response of semi-submersible[D]. Shanghai: Shanghai Jiao Tong University, 2013.
[2]	顾兴健, 赵璐, 金明, 等. 基于LSTM神经网络的我国典型试航海域环境短期预报方法研究[J]. 中国造船, 2017, 58(4): 100-107.
	GU Xingjian, ZHAO Lu, JIN Ming, et al. Research on short-term prediction of typical trial sea environment in China based on LSTM neural network[J]. Shipbuilding of China, 2017, 58(4): 100-107.
[3]	LEE S C, CHOI J Y, PARK K S, et al. Use of optical video imagery to improve wave run-up prediction accuracy[J]. Journal of Coastal Research, 2018, 85: 1271-1275. doi: 10.2112/SI85-255.1 URL
[4]	POWER H E, GHARABAGHI B, BONAKDARI H, et al. Prediction of wave runup on beaches using Gene-Expression Programming and empirical relationships[J]. Coastal Engineering, 2019, 144: 47-61. doi: 10.1016/j.coastaleng.2018.10.006 URL
[5]	BAKHTYAR R, BAKHTIARY A Y, GHAHERI A. Application of neuro-fuzzy approach in prediction of runup in swash zone[J]. Applied Ocean Research, 2008, 30(1): 17-27. doi: 10.1016/j.apor.2008.02.004 URL
[6]	ABOLFATHI S, YEGANEH-BAKHTIARY A, HAMZE-ZIABARI S M, et al. Wave runup prediction using M5' model tree algorithm[J]. Ocean Engineering, 2016, 112: 76-81. doi: 10.1016/j.oceaneng.2015.12.016 URL
[7]	BONAKDAR L, OUMERACI H, ETEMAD-SHAHIDI A. Run-up on vertical piles due to regular waves: Small-scale model tests and prediction formulae[J]. Coastal Engineering, 2016, 118: 1-11. doi: 10.1016/j.coastaleng.2016.08.008 URL
[8]	MATSUMOTO F T, WATAI R A, SIMOS A N, et al. Wave Run-up and air gap prediction for a large-volume semi-submersible platform[J]. Journal of Offshore Mechanics and Arctic Engineering, 2013, 135(1): 1-9.
[9]	李昊波, 肖龙飞, 魏汉迪, 等. 基于LSTM网络的浮式海洋平台运动在线预报研究[J]. 船舶力学, 2021, 25(5): 576-585.
	LI Haobo, XIAO Longfei, WEI Handi, et al. Research on on-line prediction of floating offshore platform motions based on LSTM network[J]. Journal of Ship Mechanics, 2021, 25(5): 576-585.
[10]	GRAVES A. Supervised sequence labelling with recurrent neural networks[M]. Berlin: Springer, 2012.
[11]	SRIVASTAVA N, HINTON G E, KRIZHEVSKY A, et al. Dropout: A simple way to prevent neural networks from overfitting[J]. Journal of Machine Learning Research, 2014, 15: 1929-1958.

序号	神经网络层	数据尺寸
1	输入层	(None, 128, 4)
2	LSTM_1	(256, 1 024)
3	LSTM_2	(256, 512)
4	Dropout	(128, 512)
5	全连接层	(128, 512)
6	输出层	(512, 1)

参数	实型值/m
总长×型宽×型深	110.0×75.0×40.5
浮箱长×宽×高	104.0×17.5×10.0
立柱长×宽	16.0×15.0
下甲板底部型高	36.5
生存吃水	18.0

工况编号	H_s/m	T_p/s	γ	随机种子数
W1	9.8	10.5	5.00	4
W2	11.8	12.2	5.00	4
W3	13.8	14.6	3.44	16
W4	15.8	15.9	3.15	16

提前预报量/s	预报精度/%
提前预报量/s	W1	W2	W3		W4		平均值
0	98.97	96.20	95.84	94.57	99.96	95.89	96.91
6	95.27	86.73	91.17	91.26	97.22	95.75	92.90
12	86.42	75.16	84.38	81.96	90.99	85.62	84.09

统计量	提前预报量/s	预报值(相对误差)/%
统计量	提前预报量/s	W1	W2	W3		W4
最大值	试验值	11.51	14.22	15.48	13.23	14.42	14.26
	0	11.76(2.1)	13.14(7.6)	15.04(2.8)	13.73(3.8)	12.89(10.6)	12.14(14.9)
	6	11.52(0.1)	12.70(10.7)	15.66(1.1)	11.81(10.7)	11.58(19.7)	11.94(16.3)
	12	10.43(9.4)	9.86(30.7)	13.72(11.4)	10.50(20.6)	10.37(28.1)	11.85(16.9)
最小值	试验值	-6.30	-6.55	-6.51	-6.81	-6.46	-6.63
	0	-6.44(2.3)	-6.49(1.0)	-6.56(0.8)	-6.69(1.7)	-6.87(6.4)	6.85(3.4)
	6	-6.47(2.6)	-6.72(2.6)	-6.68(2.6)	-6.69(1.8)	-6.70(3.7)	-6.70(1.1)
	12	-6.54(3.7)	-6.47(1.3)	-6.39(1.8)	-6.83(0.4)	-6.61(2.4)	-6.84(3.1)
有义值	试验值	3.11	2.76	2.73	2.47	2.84	2.76
	0	3.16(1.3)	2.82(2.1)	2.73(0.3)	2.48(0.4)	2.84(0.2)	2.70(2.3)
	6	2.92(6.3)	2.52(8.8)	2.57(5.8)	2.52(2.3)	2.98(5.1)	2.72(1.5)
	12	2.80(10.1)	2.42(12.2)	2.47(9.3)	2.30(6.7)	2.75(2.9)	2.44(11.7)