2.5 Development of Expert System
In order to develop an expert system for the safe operation of E-Berth, a learning based expert system was adopted. Real on-site data should be collected as many as possible for the construction of learning database, however, it was practically difficult to collect the enough number of learning data. For this reason, fast time simulations were carried out to prepare enough number of learning data bases for various situations.
2.5.1 Formulation of Expert System
To implement an expert system, input parameters should be determined first. In this research, eight parameters were selected as major factors which might affect maneuvering difficulty of berthing/ deberthing ship as described in section 2.3. Fig, 4 shows the schematic diagram for the construction of the expert system.
Construction of Expert System for the Assessment of Maneuvering Difficulty at E-Berth
The functional relationships between the maneuvering difficulty and various input parameters were sought from the learning. Input and output parameters used for learning were as follows:
Input Parameters (8 parameters)
● DWT of a vessel : real
(Ex. : 12,000, 50,000, ...)
● Berthing or Deberthing : integer
(Berthing = 1, Deberthing = 2)
● Loading Condition(%) : real
(Ballast = 0.0, Full Load = 100.0)
● Berthing Direction : integer
(Stbd. = 1, Port = 2)
● Tidal Current : integer
(1= Weak Flood, ...8=High Slack)
● Wind Speed : real
(range = 0.0 〜 35.0 knots)
● Wind Direction : real
(range = 0.0 〜 359.0, N=0.0,
● Available Tug Capacity : real (HP)
Output Parameters (1 parameter)
● Maneuvering Difficulty : real
1.0 〜 1.5 : Very Easy
1.5 〜 2.5 : Easy
2.5 〜 3.5 : Moderate
3.5 〜 4.5 : Difficult
4.5 〜 5.0 : Very Difficult
2.5.2 Verification of Expert System
To verify the output estimated by the expert system, its estimated results were compared with those by human - pilots. Fig. 5 shows the comparison results. Some discrepancies are shown between the two results, which were due to the parameters not considered in the expert system, e.g. other passing ships during berthing operations of the own ship
Comparison of Estimated Results for A-Berth Arrival Conditions
2.6 Development of GUI System
In order for the users to use the system more conveniently and to construct databases for future correlation analysis, GUI software was developed and supplied to the end user together with the various database and its management system. The software consists of eight major parts as follows:
● Expert System for the Estimation of Maneuvering Difficulty (Fig. 6)
● Data Base Management for Past Berthing/ Deberthing History Data Base
● Tidal Current Information (Fig. 7)
● Wind Information(Real time, connected to anemoscope) (Fig. 8)
● Vessel Information and related Data Base Management
● Berth Information
● Tug Information and related Data Base Management
● program Information
Fig. 6 〜 Fig. 8 show some GUI windows for this system.
Fig. 6 shows the GUI of the expert system window for the estimation of maneuvering difficulty at specific situation. It is consisted of various input windows for environmental conditions as well as vessel conditions and output window displaying the maneuvering difficulty.
Fig. 7 shows the GUI window for the tidal current information. Current velocity vector distributions around the berth can be displayed at specific time (default : present time by using the date/time information of the computer system). The variation of velocity, distribution, and water depth at specific location can be displayed also.
Fig. 6 GUI Window for the Expert System
Fig. 7 GUI Window for the Tidal Current Information
Fig. 8 GUI for the Wind Information
Fig. 8 shows the GUI window for the wind information. Time histories of wind direction and velocity are automatically displayed and stored by using the data obtained from anemoscope.
3. CONCLUDING REMARKS
The general procedures for the development of an expert system for the objective and reasonable assessment of maneuvering difficulties when berthing or deberthing are briefly reviewed together with the application results to a specific product tanker terminal. This specific expert system is currently being upgraded by using the correlation data obtained from the real on-site assessment results by human berth master.
This kind of expert system is expected to provide a berth master with objective and reasonable criteria when to prohibit the berthing or deberthing operation depending on not only the environmental conditions but also the vessel conditions.
The content of this paper is one of the results of Inherent Research Project of KRISO, "Development of Key Technologies for Maritime Risk Reduction".
 In-Young Gong & et al, "Safety Assessment of Ship Navigation by Auto-tracking Algorithm", Proc. of 1 st Korea-Japan Simulator Workshop, Jul., 2001, Korea Maritime University
 In-Young Gong and et al., "Development of Harbor Assessment Technique Using Automatic Ship Maneuvering Algorithm (In Korean)", Proc of KOSMEE Spring Annual Meeting, 1998.
AUTHOR' S BIOGRAPHY
First author, In-Young Gong, got his Ph.D. at Seoul National University in 1987 with thesis "Time Domain Analysis of Hydrodynamic Forces acting on 3 dimensional Body". Since then, he is working for KRISO in the field of ship maneuverability, development of shiphandling simulator system, and maritime traffic safety assessment.