Model-based Monitoring and Optimization of a Bio-based Process

The present thesis is concerned with establishing a scientific-based operational strategy for operations that utilize biological raw material. Bio-based processes deal with the unavoidable natural disturbances that entail the processing of biological feedstock. The final properties of a product are invariably linked with the initial properties of the raw material.

Nowadays large segments of the industry operate in a heuristic recipe-driven way, dependent on rule-of-thumb experience which too often leads to batch-to-batch discrepancies. To this end, the thesis deals with the development of tools necessary for the implementation of a flexible operational strategy.

The tools are developed to fit the rationale of incorporating critical material attributes with a desired product quality target to acquire optimized process conditions. The optimization provides the conditions at which the process should run, for the raw material with the assessed critical material attributes, to achieve the desired quality.

Furthermore, there is a need to assess whether both the defined conditions are ideal for the production system, and our predictions of product end quality are correct. Thus, a predictor correction through the incorporation of in-operation measurements is a necessary component for continuous process verification and improvement.

An industrial case study of pectin production, focused on the batch extraction from citrus fruit peels, is developed in collaboration with CP Kelco. Citrus peels are analysed to demonstrate the conceptual and performance differences of distinct quality assessment approaches. The analysis demonstrates the advantage of characterization through multivariate data analysis coupled with a complementary spectroscopic technique, near-infrared spectroscopy.

The quantitative comparative analysis of three different approaches, discriminant classification basedon expert-knowledge, unsupervised classification, and spectroscopic correlation with reference physicochemical variables, is performed in the same dataset context. A mathematical model developed by Andersen et al. (2017) is considered for optimization of the process, taking into account raw material quality uncertainty.

Before application, the model is evaluated through local sensitivity analysis. The impact of raw material uncertainty was further assessed through an uncertainty analysis, which quantified the variability of model predictions for three different types of fruit. The model provides a good agreement with the general depiction of the extraction phenomena.The critical operating parameters, i.e., temperature, pH and batch time, are then optimized in a deterministic manner to maximize the final pectin concentration while satisfying given requirements.

A robust (worst-case) optimization is examined to design the optimal operational conditions in consideration of the inherent uncertainty of feedstock and the desired product quality.The model prediction corrections are made through the use of the continuous discrete Kalman filter algorithm. A systematic approach to constructing this predictor with a desired performance is presented.

Discrepancies between the measured outputs and the filter model are observed. These differences are in the initial state guess and the considered model parameters. Implementation corrections are proposed to cope with these challenges and are evaluated for production scale data. Ultimately, the combination of these tools is showcased for a particular peel, and the impact of the proposed operational strategy is assessed.