The Application Of The Improved Rbf-nn Based On Optimization In Motor Control

RBF neural network is a kind of mathematics tool that basically doesn't rely on pattern and has strong learning ability and adaptive ability. It is suitable to have the control of uncertain and highly nonlinear systems such as electric motor. However there also exist some obvious flaws on RBF neural network. On one hand it strongly depends on the setup of initial parameters. Once the given initial parameters are error, the structure of neural network will not be optimized. On the other hand the clustering quality of RBF neural network is not high. Traditional clustering algorithms such as K-means algorithm although has fast global search speed, but it is only a rough search process and Need to explore new algorithms in order to get the global optimum search results. This text will solve the center value of RBF neural network with the interinfiltrate of ant colony optimization algorithm and chaotic ergodic theory and the traditional K-means algorithm. Then the algorithm optimized neural network is applied to the control of the electric motor to test that the method is advanced and feasible.First, the ant colony algorithm is introduced to the RBF neural network clustering algorithm, and the traditional clustering algorithm of RBF neural network is optimized using the feature that K-means algorithm has fast global search and ant colony algorithm can avoid local minimum. Using the optimized RBF neural network approximate a nonlinear function. Simulation results show that Optimized neural network has better approximating results of the nonlinear function than conventional RBF neural network, and is more conducive to the analysis and solution of nonlinear systems.Second, ergodic theory of chaotic optimization is introduced and analyzed. Logistic mapping is improved in order to meet the requirements of central value of RBF neural network to obtain. The clustering algorithm of RBF neural network is optimized using the feature that ergodicity of chaotic has fast globe search speed and can avoid local extremum.This text explores the subsequent further improvement of the ant colony algorithm, increasing the chaotic disturbance of ant colony optimization to avoid the creation of stagnation phenomenon in the calculate way. RBF neural network that optimized with chaos ergodicity and chaos ant colony is used to approximate a nonlinear function. Simulation results show that RBF neural network that optimized with chaotic ant colony algorithm has batter approximate result of the nonlinear function, which proved that adding chaotic disturbance to ant colony algorithm is feasible and practical.