Data-driven identification and modeling for cleaning of surfaces covered with film-like soils

In the food industry, processing plants are cleaned daily, and considerable amounts of water and chemicals are used. The cleaning procedures usually consume more resources than necessary since they are not optimized. Optimization may be conducted using simulations, however, currently suitable models are not available. This contribution summarizes research results of recent years in the field of cleaning modeling for the simulation-based optimization of cleaning processes at TUD. The following achievements are discussed: i) two machine learning-based strategies for classifying soils according to their behavior during removal – termed cleaning mechanism, ii) basic simulation models for each cleaning mechanism and validation of the models on various flow configurations involving a duct flow, a duct flow with sudden cross-sectional expansion, a pipe flow, and an impinging jet and iii) a combined cleaning model allowing the transition between cleaning mechanisms. The model accounts for the influence of temperature and hydroxide ion concentration of the cleaning fluid on the cleaning process. Its potential is demonstrated in a case study. In all investigated scenarios, performing cleaning simulations takes less than ten minutes. The results are placed in the context of the current state of research, and challenges for the future are identified. • Summary of results in cleaning modeling for simulation-based optimization of cleaning processes @TUD • Two ML algorithms for soil classification according to their cleaning mechanism • Simulation models for the cleaning mechanisms adhesive detachment, cohesive separation and viscous shifting • Combined cleaning model with influence of temperature and sodium hydroxide concentration • Future challenges identified