| Summary: | 碩士 === 國立臺灣大學 === 造船及海洋工程學研究所 === 89 === In this thesis, we develop a matched-filed processing (MFP) with the annealing genetic algorithm. The MFP inverts the geo-acoustic parameters by matching acoustic field data, which is recorded by a hydrophone array, with numerical replica field. While we consider the MFP as an optimization problem, it is unavoidable that a huge amount of replica field will be needed for matching with the measured field data. Furthermore, account for the complicated geo-acoustic parameters, the inversion becomes difficult to deal with. In order to obtain efficiency and accuracy, we develop a hybrid optimal algorithm, i.e. the annealing genetic algorithm.
The simulated annealing algorithm has the property of better local search ability, while the genetic algorithm might jumped over more local optimal and have the better ability in global search (but optimal solution is still not easy to obtained). By merging simulated annealing algorithm into genetic algorithm, we develop annealing genetic algorithm that provides better efficiency and is suitable for MFP. Four scenarios of ocean model are used as test case. The geo-acoustic parameters in ocean bottom are accurately estimated.
|
|---|
