Machine-Learning-Assisted Hybrid Sensor Approach for Low-Concentration Monitoring in Oil/Water Emulsions

The study of emulsions in the oil industry is essential due to their impact on oil recovery and production processes. Produced water containing oil and harmful compounds poses significant environmental risks, underscoring the need for effective monitoring and regulation. This study proposes a hybrid sensing method to measure low-concentration oil/water (O/W) emulsions with the combination of different sensing techniques including refractive index (RI), turbidity, and zeta potential, ensuring compliance with environmental standards. Techniques such as optical microscopy, refractometry, turbidimetry, zeta potential, and particle size analysis were applied to O/W systems (10–60 ppm). Turbidity measurements exhibited an exponential increase in particle count with concentration, demonstrating their effectiveness in monitoring low-concentration emulsions. In contrast, refractive index (RI) and zeta potential showed inconsistent trends, while particle size analysis revealed distinct distribution patterns at specific concentrations. Turbidimetry emerged as a reliable method for evaluating low-concentration O/W emulsions. Then, the sensor fusion approach is performed using the random forest (RF) regression algorithm in which a high accuracy is obtained with a root-mean-squared error (RMSE) of 0.2 ppm. The proposed hybrid sensor approach provides a practical solution with potential of field applications and low-cost solution for environmental monitoring in industrial applications.