| 摘要 |
PURPOSE: A system and a method for designing a ship using the neuro-fuzzy theory are provided to optimize the design of a ship by redefining respective input patterns from dimensionless input/output data, processing the redefined input patterns via a neural network, and applying the hybrid learning rule of a neuro-fuzzy system. CONSTITUTION: A system for designing a ship using the neuro-fuzzy theory comprises an information scattering unit(1) for redefining non-linear input data by identically defining the active range of a weight of dimensionless input data with the active ranges of respective neurons and scattering and storing input information, a knowledge base building unit(2) including a database(2a) defining the membership function, an information unit(2b) related with a pattern information rule, and a neural network structure recognition database(2c) defining various patterns to apply to a neural network and recognizing the structure of the neural network, a pseudo-determining unit(3) for reasoning a result by continuously referring to the knowledge base via various operations, and a reasoning unit(4) for reasoning an output value desired by the neural network from the input pattern by the pattern information rule outputted from the knowledge base building unit.
|
