Design of a Robust Soft-Sensor to Monitor In-Line a Freeze-Drying Process

Abstract

This article deals with the in-line monitoring of a vital freeze-drying process using a soft-sensor that couples the experimental measurement of product temperature with a mathematical model of the process. This tool allows estimating in-line the residual amount of ice in the product and, thus, the duration of the primary drying stage, as well as some model parameters like the heat transfer coefficient between the shelf and the product and the resistance of the dried cake to vapor flow. As the performance of this sensor, based on the Extended Kalman Filter algorithm, strongly depends on the accuracy of the initial estimations of model parameters, an innovative algorithm based on a simple model of the freezing stage (and on product temperature measurement in this stage) is first used to roughly estimate the dried cake resistance to vapor flow. Then, a curvilinear regression algorithm is used in the first part of the primary drying stage with the goal of further refining model parameters' estimations. In this way, the robustness of the sensor is significantly improved, as confirmed by various experiments carried out to validate the soft-sensor and to investigate the effect of the accuracy of the initial estimations of model parameters on the performance of the system.

Keywords:

• Freeze drying
• Mathematical modeling
• Monitoring
• Pharmaceuticals
• Soft-sensor