日本財団 図書館

共通ヘッダを読みとばす


Top > 産業 > 運輸.交通 > 成果物情報

Conference Proceedings Vol. I, II, III

 事業名 海事シミュレーションと船舶操縦に関する国際会議の開催
 団体名 日本船舶海洋工学会 注目度注目度5


PREDICTION OF THE MOTION CHARACTERISTICS OF A BARGE TRAIN USING NEURAL NETWORK RECURSIVE MODEL
Xiangjun Kong (Wuhan University of Technology, P.R. China)
Zaojian Zou (Wuhan University of Technology, P.R. China)
Junmin Mou (Wuhan University of Technology, P.R. China)
Yong Li (Wuhan University of Technology, P.R. China)
 
 Abstract: Barge train is one of the main transportation forms in Chinese inland waterways. Its hydrodynamic performance is very complicated due to a variety of combination, water depth, wind and current as well. On the basis of full-scale tests, a kind of Neural Network Recursive Model (NNRM) is established to predict the motion characteristics of a barge train with the environmental disturbances being taken into account. The simulation results have shown that the NNRM can overcome the difficulties brought by the dynamic model.
 
1. INTRODUCTION
 Yangtze River is the greatest river of China with the length of 6,300 kilometers from west to east. The shipping transportation in it plays a very important role on Chinese economic stage. Barge train is an ideal way of transportation in Yangtze River because of its more capability, lower cost and higher efficiency. However, to precisely predict the maneuvering characteristics of barge train in such inland waterways is wry difficult due to the restriction of waterways and the complicated environmental conditions.
 
 To effectively control ship motion under the complicated environmental conditions is highly desirable to know her maneuvering characteristics, which is commonly expressed by ship maneuvering equations. Nevertheless, the hydrodynamic coefficients in motion equations of the barge train are difficult to acquire since its intrinsical character and the chaotic external effects [1].
 
 In this paper, a neural network intelligent recursive model for predicting the heading, sway, and yaw motion of barge trains is proposed. This Neural Network Recursive Model (NNRM) uses ship motion state variables and rudder deflection as inputs to provide accurate prediction of the vessel's motion. The effectiveness of this method is verified by simulating full-scale tests of a certain barge train in Yangtze River, and the results are satisfactory.
 
2. MANEUVERING MOTION EQUATIONS
 Generally, a barge train consists of several barges and a push boat as shown in Fig.1. Its hydrodynamic characteristic is to a great extent determined by the combination form of the barges. Different combination form gives different characteristic.
 
2.1 The Combination of the Object Barge Train
 
 The following figure shows the object barge train used in the hill-scale tests conducted in Nantong, east China, near Shanghai. The particulars of this barge train are listed as follows:
 
Ship Name: Yangtze 22018
Length: 200m; Breadth: 65m; Mean Drift: 2.6m
 
Fig.1 The combination of the barge train.
 
2.2 Brief Introduction of the Full-scale Tests
 
 The full-scale tests are initiated by the engineering project of building a bridge across Yangtze River in Nantong. In order to represent all kinds of ships across the future bridge, the barge train with 16,000 DWTs was chosen and 4 zigzag tests were made, with serial number Test1, Test2, Test3 and Test4. Test1, Test2 and Test4 were conducted upstream, whereas Test3 downstream. The position data was recorded by bow-fixed and stem-fixed DGPS, yaw angle by gyroscope, and time by stopwatch.
 
2.3 Maneuvering Equations of Barge Train
 
 Two right-handed coordinate systems, the space-fixed coordinate system O-X0Y0Z0 and the body-fixed coordinate system G-xyz, are adopted in the description of the maneuvering motion of the barge train, as shown in Fig.2.
 
 X, Y are the components of force vector F acting on the barge train in the body-fixed system and N is the moment about z axis. r is the yaw rate and u, v are the velocity components of the speed V, φ is yaw angle and a is rudder angle.
 
Fig.2 Coordinate systems.
 
 Same as ordinary ships, the maneuvering motion equations for barge train can be written as follows:
 
 
 The terms with subscripts H, P, and R represent the hydrodynamic forces acting on hull, propeller, and rudder, respectively. The terms with subscripts NL and E represent nonlinear and external ones.
 
 The following deduction is based on the assumptions:
 
  1) Both the external and nonlinear terms can be ignored under the condition of mild rudder deflection, say within 15o.;
 
  2) Force in y direction and moment produced by propeller are negligible since they are small compared to other terms.
 
  3) Force change in x direction is small, that is, the velocity u can be considered constant.
 
2.4 Linearized Maneuvering Equations
 
 For convenience of calculation using the NNRM mentioned above, we firstly linearize the maneuvering equations and then non-dimensionalize all terms. As a result, the following equation is obtained [2]:
 
 
 Here the terms with (') denote non-dimensionalized ones. Eq. (2) is a 1st order differential equation group, with free terms of rudder force and moment. From the linearized differential equation group, the similarity between v and r is manifest.
 
3. MOTION PREDICTION BASED ON NEURAL NETWORK
 Neural Network (NN) has been widely applied to solve ship hydrodynamic problems [3][4][5]. All these efforts are mainly oriented towards sea-going ships or Autonomous Underwater Vehicles (AUVs).
 
 This section is to develop a NN recursive prediction model for simulating the maneuvering motion of barge train s in inland waterways.
 
3.1 Maneuvering Equations in Matrix Form
 
Eq. (2) is easy to be rewritten in the matrix form,
= AX + Bu , i.e.:
 
 
 With recalling the full-scale data form, it hints that the application of the proposed method calls for the use of a discrete model in time domain. In the present case:
 
 
 To be consistent with the measured data, the discrete time Δt is given as unitary time. Thus Eq. (3) can be rewritten in discrete matrix form:
 
X(k + 1) = AX(k) + Bδ(k) (4)
 
 where: x (k), X (k + 1) are discrete state vector on unitary time serial k and k+1; and
 
 
 Eq. (4) gives a recursive way to predict the state vector [v' (k); r' (k)] only if the initial condition is known; and to get the yaw angle φ (k) by integration of r' (k).







サイトに関するご意見・ご質問・お問合せ   サイトマップ   個人情報保護

日本財団会長笹川陽平ブログはこちら



ランキング
注目度とは?
成果物アクセスランキング
124位
(31,428成果物中)

成果物アクセス数
103,510

集計期間:成果物公開〜現在
更新日: 2019年8月10日

関連する他の成果物

1.MARSIM'03 開催報告
2.本会議及びワークショップ開催状況写真
3.九州における離島住民からみた交通バリアフリー化に関する調査研究?鹿児島県をモデルケースとして? 報告書
  [ 同じカテゴリの成果物 ]


アンケートにご協力
御願いします

この成果物は
お役に立ちましたか?


とても役に立った
まあまあ
普通
いまいち
全く役に立たなかった


この成果物をどのような
目的でご覧になりましたか?


レポート等の作成の
参考資料として
研究の一助として
関係者として参照した
興味があったので
間違って辿り着いただけ


ご意見・ご感想

ここで入力されたご質問・資料請求には、ご回答できません。






その他・お問い合わせ
ご質問は こちら から