Prediction of the Adsorption Amount of Azo Dyes Pollutants from Wastewater Using Porous Metal-organic Framework Adsorbents

In this project, the potential capability of intelligent machine learning methods such as LS-SVM, RBFNN, MLPNN, and ANFIS was investigated for estimating the efficiency of azo dyes removal from wastewater. To this aim, a huge data bank of azo dyes adsorption by metal-organic frameworks as porous adsorbents were collected under different conditions, including adsorbent dosage, initial dye concentration, solution pH, specific surface area, temperature, and contact time. Assessing different statistical parameters and comparing the models showed that LS-SVM approach had the minimum error and, therefore, it conferred the most accurate prediction for the efficiency of azo dyes removal by MOFs among other models. The values of AARE (%), R2</sup>, STD, and RMSE were calculated 1.844 %, 0.9899, 0.0213, and 18.511, respectively for LS-SVM. Also, this model illustrated precise compatibility with adsorption trends by variation of initial dye concentration, pH, and temperature. The sensitivity analysis presented that adsorbent surface area and adsorbent dosage had a positive impact and initial concentration and pH negatively influenced estimating the removal of dyes.</em>