Chaotic Neural Network Algorithm And Its Application To Design The Deepwater Motor

This thesis put forward a new optimal algorithm applied into thruster motor of deepwater robot.That is algorithm of chaotic neural network.Search for a large-scale parallel and intelligent features algorithm is of great significance in the actual project problem such as complicated,constrained,multi-minimum characters.The results show that the chaotic neural network algorithm is feasible to improve dynamic and static performance of the deep-sea thruster motor.In order to obtain a feasible optimum algorithm,four algorithms' theories,searching steps and features are discussed in the thesis.These algorithms are genetic algori thm(GA),simulated annealing(SA),chaos algorithm(CA) and aeural network algorithm (NNA).The advantages and disadvantages are pointed out through analysis,compariso n and verification by the optimizing test functions.Then,this paper highlighted the basic principles and characteristics of chaotic algorithms and neural network algorithms.Logistic mapping and Ulam-von Neumann mapping are analyzed systematically in thesis.The intrinsic stochastic property,ergodicity, regularity and sensitivity to initial values are shown in this paper.Because the random characteristic of chaotic motion,some times chaotic variables as close as the global optimum,but sudden jump out,so,resulting in waste of the search time.Hopfield neural network have rapid convergence for search the global optimal solution,but it easily step into a local minimum point.To deal with existing problems,combines the chaos characteristics and hopfield network characteristics,make full use of the advantages of the two algorithms in hopfield network introduced the chaotic mechanism.Introduced a chaos neural network algorithm in the design of the electrical,the algorithm has a strong ability to overcome the local minimum and the rapid convergence.This thesis used international standards multi-peak function to verify the chaotic neural network algorithm,optimization results shows that the chaotic neural network algorithms have very good capabilities of global convergence and stabilization.The feasibility and utility are confirmed by the optimum example of PMSM. Under the condition of unchanged working performance,the volume and efficiency are decreased and improved respectively.Finally,established the optimization objective function of deepwater thruster brushless motor and constrained conditions of deepwater brushless DC motor(BLDCM),the main dimension,efficiency and volume of deepwater BLDCM are optimized by using the chaotic neural network algorithms.The no-load magnetic field is analyzed by the finite element analysis software.The result of fourier decomposition for air-gap flux density proves the validity of the optimization.