SoftSensors for Process Analytical Technologies

Online monitoring of important process parameters

The SoftSensor (“Software Sensor”) serves the automatic online calculation of estimates for process parameters, which are not measurable online with conventional analytical methods, or only measurable at considerable expense. The SoftSensor is non-invasive, i.e. works without sampling. Instead it uses the measurements of already available process parameters. These measurements are being evaluated in the Soft Sensor data model and thus an online estimate for the desired process parameter is calculated (i.e. nutrient concentration or cell density).


SoftSensors as an element of modern advanced process control

  • Real time online measurement of important process parameters (i.e. cell number, optical density, nutrient concentration)
  • Increased reproducibility of the process phases (seeding, nutrient feed, induction, harvest)
  • Increase in productivity and more efficient processing by reproducibly maintaining the optimal process
  • Facilitates intelligent advanced process control and process analytical technologies (PAT)
  • Early detection of faults and discrepancies
  • No maintenance work necessary, no regular calibration required
  • Cost reduction for sampling and laboratory as manual samples and offline analyses can be eliminated

Proven technology for different applications

SoftSensor is a software which turns incoming measurements into estimates of unmeasured process parameters with the help of a data model. These process parameters would otherwise not be measurable at all, or very costly to measure. These estimates can then be used for monitoring or controlling, just as if they came from a real analytical device.

The data model in SoftSensor contains a mathematical description of the correlation between the measured inputs and the process parameters (outputs) sought, and can be set up for any application. The SoftSensor operates simultaneously to the plant and can deliver up-to-date estimates for the desired process parameters in intervals of seconds or minutes.

For the calibration of a SoftSensor historic data with measurements of the respectively relevant process parameters is evaluated. Only these initial data records have to contain (offline) measurements of the process parameters to be estimated. Subsequently these measurements are no longer required. Depending on the type of application three sets of measurements can suffice to set up the data model in SoftSensor.

Details on linking SoftSensors to the process are available here .