A machine learning approach for estimating surface tension based on pendant drop images

An image-based machine learning (ML) method is introduced to predict the surface tension of an ethanol-water pendant drop made unknown composition. In contrast previous neural-network based predictors that rely on using either simplified molecular-input line-entry system or experimentally measured values, present deep neural network model directly predicts tensions arbitrary shapes at any stage before breakup. Using convolutional networks (CNNs), values are accurately obtained independent size and liquid properties. Two CNN architectures presented a for three different ML models. To improve generality models, image data augmentation technique used generate more representatives from available data. Approximating drops outside range given classes has also been demonstrated. The trained models have overall accuracy about 98% in predicting containing Additionally, tested methanol-water mixtures demonstrate results show good accuracy. Besides performance measures including Precision, Recall F1-score reported each value dataset. Compared with physics-based axisymmetric shape analysis techniques, not limited equilibrium images much faster versatile. This shows great promise vastly images.