Tehran University of Medical Sciences

Science Communicator Platform

Stay connected! Follow us on X network (Twitter):
Share this content! On (X network) By
Machine Learning Algorithm and Neural Network Architecture for Optimization of Pharmaceutical and Drug Manufacturing Industrial Effluent Treatment Using Activated Carbon Derived From Breadfruit (Treculia Africana) Publisher



Ovuoraye PE1, 2 ; Ugonabo VI1 ; Fetahi E3 ; Chowdhury A4 ; Tahir MA5 ; Igwegbe CA1, 6 ; Dehghani MH7, 8, 9
Authors
Show Affiliations
Authors Affiliations
  1. 1. Department of Chemical Engineering, Nnamdi Azikiwe University, P.M.B. 5025, Awka, 420218, Nigeria
  2. 2. Department of Chemical Engineering, Federal University of Petroleum Resources, P.M.B. 1221, Effurun, Nigeria
  3. 3. Faculty of Computer Science, University of Prizren “Ukshin Hoti�, Prizren, Kosovo
  4. 4. Technical Services, LLC, Dayton, NJ, United States
  5. 5. Department of Multimedia, UFR STGI-Universite de Franche-Comte, 4 PI, Lucien Tharradin, Montbeliard, 25200, France
  6. 6. Department of Applied Bioeconomy, Wroclaw University of Environmental and Life Sciences, Wroclaw, 51-630, Poland
  7. 7. Department of Environmental Health Engineering, School of Public Health, Tehran University of Medical Sciences, Tehran, Iran
  8. 8. Center for Water Quality Research, Institute for Environmental Research, Tehran University of Medical Sciences, Tehran, Iran
  9. 9. Center for Solid Waste Research, Institute for Environmental Research, Tehran University of Medical Sciences, Tehran, Iran

Source: Journal of Engineering and Applied Science Published:2023


Abstract

In a recent development, attention has shifted to the application of artificial intelligence for the optimization of wastewater treatment processes. This research compared the performances of the machine learning (ML) model: random forest, decision tree, support vector machine, artificial neural network, convolutional neural network, long-short term memory, and multiple linear regressors for optimization in effluent treatment. The training, testing, and validation datasets were obtained via the design of an experiment conducted on the removal of total dissolved solids (TDS) from pharmaceutical effluent. The breadfruit-activated carbon (BFAC) adsorbent was characterized using scanning electron microscopy and X-ray diffraction techniques. The predictive capacity of an ML algorithm, and neural network architecture implemented to optimize the treatment process using statistical metrics. The results showed that MSE ≤ 1.68, MAE ≤ 0.95, and predicted-R2 ≥ 0.9035 were recorded across all ML. The ML output with minimum error functions that satisfied the criterion for clean discharge was adopted. The predicted optimum conditions correspond to BFAC dosage, contact time, particle size, and pH of 2.5 mg/L, 10 min, 0.60 mm, and 6, respectively. The optimum transcends to a reduction in TDS concentration from 450 mg/L to a residual ≤ 40 mg/L and corresponds to 90% removal efficiency, indicating ± 1.01 standard deviation from the actual observation practicable. The findings established the ML model outperformed the neural network architecture and affirmed validation for the optimization of the adsorption treatment in the pharmaceutical effluent domain. Results demonstrated the reliability of the selected ML algorithm and the feasibility of BFAC for use in broad-scale effluent treatment. © 2023, The Author(s).