Mixing Performance Prediction of Detergent Mixing Process Based on the Discrete Element Method and Machine Learning

Abstract

The DIY approach promotes small-scale digital manufacturing for the production of customized, fast moving consumer goods, including powder detergent. In this context, a machine was developed to manufacture a customized detergent according to the needs of the clients indicated on a digital platform connected to the machine. The detergent is produced by a mixing process of the formulation components carried out in a 3D mixer. Analysing the mixing performance of the process is essential to obtain a quality product. In this study, the mixing process of the powder detergent was modelled using the discrete element method. After validating it with experimental test, this model was utilized to study the mixing performance considering the allowable mass fraction range of every formulation component and a mixer speed of 45 rpm, and the dataset generated from this study was employed along with a machine learning algorithm to obtain a model to predict the mixing index. In this sense, twenty-five different combinations of the defined components were simulated and a mixing index of 0.98–0.99 was obtained in a time of 60 s, revealing that all the combinations were completely mixed. In addition, the developed model was validated with results obtained from the DEM model. The model predicts the mixing index in advance and with accuracy.