Powder stickiness in milk drying: uncertainty and sensitivity analysis for process understanding

A powder stickiness model based in the glass transition temperature (Gordon – Taylor equations) was built for a production scale milk drying process (including a spray chamber, and internal/external fluid beds). To help process understanding, the model was subjected to sensitivity analysis (SA) of inputs/parameters, and uncertainty analysis (UA) to estimate confidence intervals on model predictions. For SA, a differential local and also a global approach were used. A variance decomposition method (e.g. Sobol first order sensitivity index) was implemented for global SA, and Monte Carlo technique for nonlinear error propagation was selected as the main UA approach. SA results show an important local sensitivity on the spray dryer, but at the end of the internal fluid bed (critical point for stickiness) minor local sensitivities were observed. Feed concentrate moisture was found as the input with major global sensitivity on the glass transition temperature at the critical point, so it could represent a key variable for helping on stickiness control. UA results show the major model predictions uncertainty on the spray dryer, but it does not represent a stickiness issue since the product is not in contact with dryer walls. The confidence interval for the glass transition temperature at the critical point was evaluated and this will be useful for further optimization efforts under stickiness constraints.