Prediction of Emulsion Stability via a Neural Network-Based Mapping Technique

Abstract

Stability is a key property related to the production and use of a number of products that are commercialized in the form of emulsions. However, because of the complexity of these systems, in most cases there is no readily available means to predict emulsion stability. Thus, product formulation development is based on laboratory evaluations of emulsion stability, e.g., by measuring the time required for the emulsion to break according to standardized procedures. These tests are time-consuming and subject to visual inaccuracies between different operators. In this work, a neural network-based model is tested as a tool for predicting the emulsification properties of mixtures of surfactants, organic solvents, and organic compounds used as active ingredients in pesticide formulations. The model is able to predict the emulsion-breaking height, a standard measure of the system's stability. Six physicochemical properties were used as descriptors for the active ingredients: octanol/water partition coefficient, molar volume, refractive index, density, Hildebrand solubility parameter, and Henry's constant. The solvents and their mixtures were described by their density, Hildebrand solubility parameter, and surface tension. Other input variables include water hardness and the concentrations of active ingredient and surfactant. The output variable was the volume percentage of cream formation after dispersion of the emulsifiable concentrate in water, represented by the “emulsion-breaking height”, predicted by the neural network model. In a different approach, nonionic surfactants were described by their average hydrophilic−lipophilic balance (HLB) number. The loss of information in the neural network resulted in inaccurate estimates of cream formation. Nevertheless, with this approach, the neural network was able to discriminate between regions of emulsion stability and instability. Application of the method enabled the construction of phase diagrams for the prediction of the optimum surfactant mixture(s) for emulsification of a given combination of solvent and active ingredient. The method can reduce the time needed for formulation development, minimize the exposure of the formulator to chemicals, and avoid unnecessary production of effluent in product test facilities.