The adsorptive removal of as (III) using biomass of arsenic resistant Bacillus thuringiensis strain WS3: characteristics and modelling studies
Globally, the contamination of water with arsenic is a serious health issue. Recently, several researches have endorsed the efficiency of biomass to remove As (III) via adsorption process, which is distinguished by its low cost and easy technique in comparison with conventional solutions. In the pre...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Published: |
Elsevier Inc.
2019
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/89212/ http://dx.doi.org/10.1016/j.ecoenv.2019.01.067 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Globally, the contamination of water with arsenic is a serious health issue. Recently, several researches have endorsed the efficiency of biomass to remove As (III) via adsorption process, which is distinguished by its low cost and easy technique in comparison with conventional solutions. In the present work, biomass was prepared from indigenous Bacillus thuringiensis strain WS3 and was evaluated to remove As (III) from aqueous solution under different contact time, temperature, pH, As (III) concentrations and adsorbent dosages, both experimentally and theoretically. Subsequently, optimal conditions for As (III) removal were found; 6 (ppm) As (III) concentration at 37 °C, pH 7, six hours of contact time and 0.50 mg/ml of biomass dosage. The maximal As (III) loading capacity was determined as 10.94 mg/g. The equilibrium adsorption was simulated via the Langmuir isotherm model, which provided a better fitting than the Freundlich model. In addition, FESEM-EDX showed a significant change in the morphological characteristic of the biomass following As (III) adsorption. 128 batch experimental data were taken into account to create an artificial neural network (ANN) model that mimicked the human brain function. 5-7-1 neurons were in the input, hidden and output layers respectively. The batch data was reserved for training (75%), testing (10%) and validation process (15%). The relationship between the predicted output vector and experimental data offered a high degree of correlation (R 2 = 0.9959) and mean squared error (MSE; 0.3462). The predicted output of the proposed model showed a good agreement with the batch work with reasonable accuracy. |
|---|