Intelligent Algorithm-Based Modelling Wastewater Treatment

Published: 30 October 2025| Version 1 | DOI: 10.17632/xn75m8s9cn.1
Contributors:
BISWANATH MAHANTY,

Description

The experimental dataset used are of two bioreactor studies, Airlift bioreactor (ALR) and Inverse Fluidized Bed Bioreactor (IFBB). Influent selenite concentration (ICselenite) and hydraulic retention time (HRT) are predictor for both datasets. The Selenite removal efficiency (REselenite) (%) is the response in ALR, while REselenite (%) and chemical oxygen demand removal efficiency (RECOD) (%) in IFBB. The datasets were trained using three conventional (Levenberg-Marquardt, Bayesian regularization, scaled conjugate gradient) and three evolutionary (genetic algorithm, particle swarm, reptile search) algorithms. In this submission we have the compiled data, selecting architecture, and graphical representation representation.

Files

Steps to reproduce

"corrmat.mat" files have two table variables IFBR, ALR for the predictors and response (measured). "coorelation_analysis.m" takes the input from "corrmat.mat", and plot the pair-wise coorelation, and produce heatmap. "Basic_analysis.m" loads exp_predict.mat containing variables "exp_pred_ALR", "exp_pred_IFBR_SE", "exp_pred_IFBR_COD", and "nodes_ALR nodes_IFBR". The experimental/predicted data is used to generate regression plot for all the responses in reactor system. "nodes_ALR nodes_IFBR" maps number of hiddenlayer neurons versus fitness across different training algorithm. The script makes fitness evolution plot for the same

Institutions

  • Karunya University
  • VIT University

Categories

Artificial Neural Network, Genetic Algorithm, Searching Algorithm, Correlation Analysis, Fluidized Bed Reactor

Licence