Skip to Content

PAC (K-Project)


Process Analytical Chemistry - Data Acquisition and Data-Processing

This project aims at developing enhanced quantification and prediction models for 2-dimensional (NIR,MIR,QCL) and 3-dimensional hyper-spectral data.
The project is coordinated by Recendt GmbH (Research Center for Non-Destructive Testing) and divided into 4 internal multi-firm projects and 2 strategic research projects:

  • MP1: Quantification of Process Gases
  • MP2: Quantifying and Predicting Parameters of Liquids in Batch Processes
  • MP3: Quantification of Parameters and Detection of Anomalies critical Parameters in Liquids within continuous Processes 
  • MP4: Monitoring the Production of Viscose Fibres
  • SP1: Advanced Chemometric Modeling
  • SP2: QCL-WAGS Sensor Systems

A specific part of this project (where our department plays a major role as key researcher) is dedicated to the strategic research project SP1, where incremental and evolving modelling methods will be investigated and further developed for adapting quantification and prediction models dynamically on-the-fly with new incoming samples. Active and semi-supervised learning techniques as well as fast segmentation algorithms for 3D hyper-spectral data will play an important role for achieving improved dynamic chemometric models. Outcomes of this research will be applied to concrete chemometric modeling/supervision scenarios at the industrial partners.
Topics: chemometric modeling, calibration, quantification, prediction, anomaly detection, output diagnostics, cost reduction for measurements, transfer learning
Applied methodologies: non-linear calibration models, incremental machine learning techniques, active and semi-supervised learning, evolving (fuzzy) systems, wavelength selection, genetic algorithms, statistics

Current Switch: