Enhancing Nanocomposite Filtration Membranes: Refined SVM Approach for Precise Estimation of Permeate Flux and Foulant Rejection

Authors P.M. Yawalkar1, P. William2 , V.M. Tidake3, P.M. Patare4, P.B. Khatkale5, , A.A. Khatri6, S.S. Ingle4
Affiliations

1Department of Computer Engineering, MET's Institute of Engineering, Nashik, India

2Department of Information Technology, Sanjivani College Engineering, Kopargaon, SPPU, Pune, India

3Department of MBA, Sanjivani College of Engineering, Kopargaon, SPPU, Pune, India

4Department of Mechanical Engineering, Sanjivani College of Engineering Kopargaon, SPPU, Pune, India

5Sanjivani University, Kopargaon, MH, India

6Department of Computer Engineering, Jaihind College of Engineering, Kuran, SPPU, Pune, MH, India

Е-mail praveenkhatkale@gmail.com
Issue Volume 16, Year 2024, Number 3
Dates Received 17 April 2024; revised manuscript received 23 June 2024; published online 28 June 2024
Citation P.M. Yawalkar, P. William, V.M. Tidake, et al., J. Nano- Electron. Phys. 16 No 3, 03016 (2024)
DOI https://doi.org/10.21272/jnep.16(3).03016
PACS Number(s) 74.25.Qt, 78.67.Sc
Keywords Thin-film nanocomposite (TFN), Machine learning, Permeate flux, Foulant rejection, Refined Support Vector Machine (RSVM).
Annotation

The nanocomposite filtration membranes have emerged as potential water purification and separation technologies. However, reliable estimation of foulant rejection and permeate flux remains difficult due to the complicated interaction of many components. Traditional modeling techniques fail to capture the complex dynamics at work. In this paper, we provide a Refined Support Vector Machine (RSVM) strategy to solve this issue and increase the performance of nanocomposite filtration membranes. To normalize the features, the data are pre-processed using min-max normalization. Data features like foulant rejection rates, permeate flux values, membrane features, and experimental setup are displayed. Furthermore, the proposed RSVM to determine the best input factors for the effectiveness of each nanocomposite membrane. Due to the strong resilience of RSVM and the great generalization ability of the ML model, the obtained results demonstrated that the RSVM model's prediction efficiency (R2 = 0.995) outperformed the mathematical model in terms of prediction performance. To conduct training, validation and testing for this work, we employed statistical data including 764 samples of the input variables (five) and output variables (two). The RSVM approach provides a dependable and effective way to forecast membrane fouling and water filtration by predicting foulant rejection and permeate flux.

List of References