Abstract:This paper studies passivity analysis and stability analysis for neural networks with time-variant delay. The time-variant delay is presumed to respond in trigonometric form. Firstly, a novel interval approximation method is constructed with adjustable parameters, which can be adjusted to reduce conservatism. According to the feature of time delay in trigonometric form, the specific allowable delay set can be obtained. Secondly, by introducing the delay-product-type parts in Lyapunov-Krasovskii functional (LKF), the emphatically improved passivity criterion and stability criterion with less conservatism are obtained by using the advanced inequalities. Finally, two numerical simulations are listed where our work can obtain larger maximum allowable delay bounds compared with other works, which demonstrate that the results based on the interval approximation method are less conservative.

A novel interval approximation method for passivity and stability analysis of delayed neural networks