Knowledge Discovery of Hydrocyclone s Circuit Based on SONFIS and SORST

TL;DR

This paper introduces a novel approach combining SOM, NFIS, and RST to analyze hydrocyclone performance, enabling improved granulation and performance prediction through approximate reasoning methods.

Contribution

It presents a new integrated method using SONFIS and SORST for hydrocyclone analysis, combining fuzzy and rough set theories with self-organizing maps.

Findings

Effective granulation of data achieved

Improved hydrocyclone performance analysis

Validated methods on hydrocyclone data set

Abstract

This study describes application of some approximate reasoning methods to analysis of hydrocyclone performance. In this manner, using a combining of Self Organizing Map (SOM), Neuro-Fuzzy Inference System (NFIS)-SONFIS- and Rough Set Theory (RST)-SORST-crisp and fuzzy granules are obtained. Balancing of crisp granules and non-crisp granules can be implemented in close-open iteration. Using different criteria and based on granulation level balance point (interval) or a pseudo-balance point is estimated. Validation of the proposed methods, on the data set of the hydrocyclone is rendered.