2002 年 2002 巻 238 号 p. 238_1-238_7
This paper describes recently developed general-purpose optimization system, in which users are able to set up the optimization problem without knowledge of system coding. Currently two optimization algorithms are implemented, i.e., the sequential quadratic programming (SQP) and the genetic algorithm (GA), both of which are the most widely used nonlinear programming techniques. The distributed-object technique, i.e., CORBA, is implemented in the present system for remote networking computation, which enables the network distributed processing to evaluate objective and the constraint functions at different computer sites. Users can set up variety of optimization problems by using spreadsheet screen.
For demonstration of the present system, two design optimization problems for container ship are considered, i.e., minimization of annual fuel consumption, and maximization of annual gross margin. Results from two optimization algorithms, i.e., SQP and GA are compared for the computational performance. Also, computational time is compared among three methods to evaluate objective and constraint functions, i.e., method using EXE and DLL files and CORBA protocol. In conclusion, the present system has been shown very promising, and can be a practical and useful design tool through further extension for CAD and CFD interfaces, which is currently in progress.