Toward Fuzzy Activation Function Activated Zeroing Neural Network for Currents Computing

In order to improve the convergence and noise resistance ability of the ZNN models, a fuzzy activation function (FAF) is designed. Based on the FAF, a fuzzy activation function activated zeroing neural network (FAFZNN) for online fast computing circuit currents is proposed. By introducing the fuzzy logic technique, the convergence and noise resistance ability of the proposed FAFZNN model are further promoted, and it realizes prescribed-time stable, which is irrelevant to its system initial states even in noisy environment. Moreover, the prescribed-time convergence and strong robustness to noises of the proposed FAFZNN model are verified by strict mathematical analysis. The comparable simulation results for static direct currents (DC) and dynamic alternating currents (AC) computing in noiseless and noisy environment further validates its superior effectiveness and robustness for practical applications.

Automated summary

This paper proposes a method to improve the convergence and noise resistance ability of ZNN models by designing a fuzzy activation function, which is applied to online fast computing of circuit currents. By introducing fuzzy logic technique, the convergence and noise resistance ability of the model are further enhanced, and prescribed-time stability is achieved irrelevant to the initial states even in noisy environment. Mathematical analysis and simulation results verify the robustness and effectiveness of the proposed method for practical applications.