Development of an Intelligent Online System for Predicting Membrane Fouling Using Real-Time TSS and SDI Estimation

Khadidja Benyahia, Naoual Bensaad, Farid Hammou

Membrane fouling severely limits water filtration system performance and lifespan and increases operational costs. Lab-based monitoring of Total Suspended Solids (TSS) and Silt Density Index (SDI) is slow, labor-intensive, and unsuitable for continuous control. Water treatment facilities need rapid, onsite, energyefficient monitoring tools.This study investigates whether an embedded, low-cost online sensing system can reliably estimate TSS and SDI in real time and serve as a predictive tool for membrane fouling risk, while reducing reliance on conventional laboratory protocols.Methodology: An intelligent prototype was developed using an embedded architecture comprising an Arduino UNO microcontroller, a turbidity sensor, an analog-todigital communication module, an LCD interface, a battery power supply, and an RS-485/TTL converter for data transmission. The system continuously acquires turbidity-related signals, which are mathematically processed using linear correlation and regression to estimate TSS and SDI values. Experimental validation was performed by comparing online measurements with reference laboratory methods used for suspended solids and SDI determination.Results: The proposed system closely matched standard analytical techniques. TSS estimation had a Pearson correlation coefficient of 0.9191 and a relative error of 5.3760%, confirming reliable prediction. Firstorder polynomial regression gave a very low root mean square error (RMSE = 7.8505 × 10⁻¹⁶), showing high model accuracy. For SDI estimation, the system had a Pearson correlation coefficient of 0.8673, a coefficient of determination (R²) of 0.7523, and a mean relative error of 2.2931%, demonstrating its ability to assess particulate fouling potential. The findings show that embedded, real-time monitoring can feasibly replace traditional laboratory analyses. Strong statistical correlations demonstrate that turbidity-based sensing and data-driven modeling reliably estimate TSS and SDI. The system’s portability, low energy use, and simple design enable effective continuous field deployment. Conclusion: The developed online preventive system provides a practical, accurate, and resource-efficient alternative for monitoring fouling-related parameters. By enabling real-time prediction of membrane fouling risk, the approach supports proactive process control and reduces operational costs. This contributes to the sustainable management of membrane-based water treatment systems.