Методы идентификации нелинейных систем с использованием искусственного интеллекта
Аннотация
В работе, в качестве нелинейных динамических моделей были исследованы две хаотические системы, осциллятор Дуффинга и модель Лоренца. Моделирование этих систем проведено в среде МАТЛАБ, с использованием решателей для жестких систем дифференциальных уравнений. Были исследованы автоколебательные процессы для неавтономного осциллятора Дуффинга с жестким режимом возбуждения. Для автономной системы Лоренца, изначально были взяты параметры, при которых в системе реализуется хаотический режим. Метод идентификации и управления для рассматриваемых динамических систем, основан на нейронной сети радиально базисных функций. В качестве программной среды для реализации предложенного метода, в работе также выбрана система МАТЛАБ.
In this paper, two chaotic systems, the Duffing oscillator and the Lorentz model, were studied as nonlinear dynamic models. The simulation of these systems was performed with MATLAB, using solvers for rigid systems of differential equations. Self-oscillatory processes were investigated for a non-autonomous Duffing oscillator with a rigid excitation mode. For the autonomous Lorentz system, the parameters at which the chaotic mode is implemented in the system were initially taken. The method of the identification and control for the considered dynamic systems is based on a radial basis function network. The MATLAB was also chosen as a software environment for the implementation of the proposed method.
In this paper, two chaotic systems, the Duffing oscillator and the Lorentz model, were studied as nonlinear dynamic models. The simulation of these systems was performed with MATLAB, using solvers for rigid systems of differential equations. Self-oscillatory processes were investigated for a non-autonomous Duffing oscillator with a rigid excitation mode. For the autonomous Lorentz system, the parameters at which the chaotic mode is implemented in the system were initially taken. The method of the identification and control for the considered dynamic systems is based on a radial basis function network. The MATLAB was also chosen as a software environment for the implementation of the proposed method.