Research On Firefly Algorithm And Its Application In Parameter Identification Of Ship Motions

Hydrodynamic parameters are very important during the aircraft design process,while some hydrodynamic parameters with high sensitivity are hard to obtain by real cruising or tank experiment,and there is a great difference between theoretical calculating and real cruising.Establishing the precise mathematical model descripting craft hydrodynamic parameters and motion states by advanced identification technology is of great significant for the craft optimization design and performance improvement.The problem of hydrodynamic parameters identification can boil down to a complex function optimization problem after determining the identification criteria.Firefly algorithm was proposed in the early 21^st century as a new intelligent optimization algorithm.It has attracted attentions from researchers home and abroad with advantages of simple structure,easy implementation,less parameters needed to be adjusted,and superiority in solving practical engineering problems and function optimizations.This paper focuses on the basic theory of firefly algorithm and its applications in the ship longitudinal motion hydrodynamic parameters identification.The main works are as follows:1.The theoretical study of mathematical model,parameters setting mechanism,time complexity and algorithm validity of the firefly algorithm?FA?has been carried out in depth.Since the FA is a stochastic search algorithm,the introduction of random mechanism not only improves the convergence rate and accuracy of the algorithm,but also brings a huge challenge to its theoretical research,which has been rarely studied compared to its pervasive application.In this paper,the basic theory of FA is deeply researched,and the parameters choosing principle is given.2.The influence of random mechanism on the algorithm performance is studied,and two improved FAs are proposed based on Lévy flight and random distribution strategy respectively.Random mechanism is the soul of intelligent optimization algorithm and has significant effect on algorithm performance.Simulation results show that the proper selection of random mechanism can help the algorithm jump out of local optimum effectively and improve the precision and convergence rate.Otherwise,it would lead the algorithm to the stagnation too early and obtaining poor optimization results even unusable.3.The multiple algorithm fusion mechanism of intelligence optimization algorithm is studied,and a hybrid FA with synchronous parallel structure is proposed.FA and differential evolution?DE?are executed in parallel during the iterative process,and the obtained population will be mixed and regrouped again.Experimental results illustrate that FA can provide an accelerated convergence rate and an increased global exploration ability,moreover DE can improve the precision and increase local detection ability.Besides,the mix and regroup mechanism can enhance the communication between populations and increase the diversity of the algorithm to avoid stagnation.4.The influence of the balance between global exploration ability and local exploitation ability on algorithm performance is studied,and a self-adaptive hybrid firefly algorithm?SaHFA?based on self-adaptive parameter control mechanism is proposed.It is proved that the balance between global exploration and local exploitation can significantly improve algorithm performance.SaHFA can adjust the parameters adaptively based on HFA,and implicit the parameters adjustment and selection in the iteration process.Experimental results show that this mechanism can adjust and balance the global exploration and local exploitation ability adaptively during runtime,as well as improving the convergence rate,accuracy and robustness.5.The ship longitudinal motion parameters identification problem is studied by using SaHFA.First of all,the discrete state equation of ship longitudinal motion is established;second,rudder angle disturbance model and wave disturbing force and torque model are built by modeling the rudder and wave disturbance respectively;third,least square error criterion is chosen as fitness function,and the ideas and implementation process of the improved FA are illustrated in detail;Forth,ship longitudinal motion parameters are identified under rudder and wave disturbance respectively,the influence of observation noise on algorithm performance is discussed,too.Finally,the simulation results are evaluated,which show that the improved algorithm can successfully identify eight parameters in ship longitudinal motion at the same time,and the precision is quite high.