The cognitive sensor solution CPS3

The cognitive sensor solution CPS3, used to predict the granule moisture in a fluid bed dryer from available process data, has been successfully tested on RCPE’s GEA ConsiGma CTL25 manufacturing line. The underlying mathematical model for estimating the granule moisture has been developed. Its parameters have been selected based on identification experiments, and its estimation quality has been validated through dedicated validation experiments.

This cognitive sensor will also be part of the cognitive control concept developed in CAPRI: The knowledge of the granule moisture will be used to actively adjust the drying time in the fluid bed dryer. Figure 1 gives an overview of the concept. The soft-sensor block uses the mathematical model to estimate the granule moisture. A comparison between measured (LODLHP) and estimated (LODsoft-sensor) granule moisture (LOD) is shown. The top right-hand block demonstrates the fundamental premise of our control concept: The difference between conventional operations with constant drying time (noctrl, blue) and our proposed solution with actively adjusted drying time based on real-time data (ctrl, red). In the case of standard operation (noctrl, blue), the granule moisture at the end of the drying cycle is too low, whereas when using the estimated value of the LOD for stopping the drying at 600s (ctrl, red), the granule moisture hits the desired set-point. The ease of implementation and compatibility with existing manufacturing lines are significant advantages, as the sensor is solely based on executing an algorithm, and no additional hardware modifications are required.

Figure 1: Use of the granule moisture sensor CPS3 as part of a process control concept. Based on the granule moisture (expressed in “loss on drying”, LOD), the drying time is adjusted such that the final granule moisture hits the desired set-point.