A novel negative-correlation redundancy evolutionary framework based on stochastic ranking for fault-tolerant design of analog circuit

Abstract

The fault-tolerant evolutionary design based on negative-correlation redundancy technique is an effective way to improve the fault-tolerance of analog circuits with uncertain faults. In the existing negative-correlation redundancy evolutionary framework (ENCF), the negative-correlation penalty coefficient plays an important role, and it affects the performance of ENCF greatly. However, the value of the negative-correlation penalty coefficient is heavily dependent on the experience of designers. In this paper, we propose a new negative-correlation redundancy evolutionary framework based on stochastic ranking strategy. In order to make comparisons with the existing researches, we employ analog filter as a design example. Experimental results show that the framework proposed in this paper can generate negatively correlated redundancies without specifying the penalty coefficient, and it shows a relatively high ability to convergence compared to ENCF. © 2013 Springer-Verlag Berlin Heidelberg.