ANFIS deriving from jointed input-output data space and applying in smart-damper identification

Abstract

For approximation of unknown mapping f: X→Y expressing a given database via an adaptive Neuro-Fuzzy inference system (ANFIS), ANFIS’s convergent capability is quite sensitive to the data features. In order to deal with this, this paper focus on ameliorating quality of cluster data space (CDS) used to establish the ANFIS. Firstly, we formulate and prove CDS-related necessary conditions for an approximation expressing an initial data space (IDS) convergent. Based on this theory basis, we propose a fuzzy system typed ANFIS associated with two solutions for establishing the CDS from the IDS, which focus on preventing, seeking and exterminating critical data samples in the CDS. In order to deploy these, we also present an improved structure of ANFIS. These aspects are described via a novel offline identification algorithm named ANFIS-JS for building ANFIS in a jointed input-output data space (JDS) deriving from the IDS. The results obtained via several surveys, including identifying smart dampers, magnetorheological damper (MRD) and electrorheological damper (ERD), show that the convergent stability and response accuracy are the main advantages of the ANFIS-JS.