ANFIS Identification applied to a reservoir level liquid system

This work applies a methodology for identification of a double tank system with conic shape by using an Adaptive-network-based fuzzy inference system (ANFIS). Considering that the cross section of the tank is no longer constant, it rather changes with the height of the liquid, a nonlinear modelling...

Full description

Saved in:

Bibliographic Details
Published in:	2021 9th International Conference on Control, Mechatronics and Automation (ICCMA) pp. 135 - 140
Main Authors:	Vasconcelos, Felipe J. S., Leite, Gabriel C., Neto, Guilherme B. F., Correia, Wilkley B., Aguiar, Victor P. B., Paiva, Davi A.
Format:	Conference Proceeding
Language:	English
Published:	IEEE 11-11-2021
Subjects:	Adaptation models Geometry Liquids Mathematical model Mechatronics Neuro-fuzzy Nonlinear identification Process control Reservoirs Shape System identification Tank system
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	This work applies a methodology for identification of a double tank system with conic shape by using an Adaptive-network-based fuzzy inference system (ANFIS). Considering that the cross section of the tank is no longer constant, it rather changes with the height of the liquid, a nonlinear modelling method becomes more suitable. A comparison is made with the phenomenological model, which shows that such approach does not show appropriate adherence to the data collected from the system, even for the nonlinear model obtained from geometry. ANFIS on the other hand, presents a considerable improved fitness as it captures far more modes of the system, being tested for five different regression vector structures and for three evaluation criteria, namely root mean square error (RMSE), the adjusted determination coefficient R^{2}_{a\, j} and Akaike Information Criterion (AIC). A trade off between complexity and fitness was used to choose the ANFIS model 3 with {RMSE}=0.0702, R^{2}_{aj} \quad =0.7135 and {AIC}=2147.2974 for the validation collected data. The findings show the effectiveness of the presented methodology herein.
DOI:	10.1109/ICCMA54375.2021.9646226