Real-time monitoring of milk powder moisture content during drying in a spouted bed dryer using a hybrid neural soft sensor

The direct measurement of the moisture content of dried products would be more interesting for process control purposes. However, the most common procedures for such measurement are either slow or expensive for industrial dryers. Alternatively, one might reduce the cost of an effective measurement procedure by using other sensors (which are less expensive and whose response is faster), which can provide information for a physical-mathematical model representing well the drying process. In this context, the objective of this work was the application of a previously developed soft sensor for the online measurement of milk powder produced in a spouted bed dryer. A hybrid neural model was used as part of a soft sensor and coupled to the data acquisition interface. The sensor was capable of estimating milk powder moisture content when the dryer was submitted to disturbances on air inlet temperature and paste inlet flow rate. On the other hand, the model failed to describe paste accumulation within the bed, which is the reason why the soft sensor tended to overestimate moisture content for longer operation times.