Call for partners | PPP INSIGHTS: In-line insight into feedstock and side-stream composition

Efficient utilisation of fibrous feedstocks and industrial side streams depends heavily on their biochemical composition, which determines both processing behaviour and final application potential. However, these materials often exhibit substantial variability due to variation in seasonal effects and batch to batch variations. Such fluctuations directly impact fractionation efficiency, energy consumption, nutritional value, digestibility, and overall product quality. Although traditional wet‑chemical analytical methods provide accurate compositional data, they are labour‑intensive, slow, and unsuitable for the real‑time monitoring required for modern process control. As a result, industries lack timely insights into changing material properties, limiting their ability to optimise operations and maintain consistent product performance.

This project addresses these limitations by developing a fast, scalable, and robust analytical framework that enables continuous inline monitoring of fibre-rich materials. The key challenge lies in the complexity of heterogeneous fibre matrices, which require simultaneous quantification of components such as starch, protein, fat, cellulose, hemicellulose, and lignin. In addition, industrial environments introduce variability that demands highly stable calibrations, sensitivity validation, and ongoing model maintenance. Integrating such analytical capabilities into automated control systems also requires advanced modelling approaches capable of handling uncertainty while ensuring safe and reliable factory operation.

To meet these needs, the project proposes an integrated monitoring and control system that combines inline, online, and at‑line FT‑IR/MIR and NIR spectroscopy with robust calibration against reference methods. Spectral data will be embedded within a hybrid modelling architecture that merges first‑principles knowledge with data-driven learning for adaptive optimisation of processing conditions. A Model Predictive Control (MPC) layer enhanced with reinforcement learning, likely supported by human oversight, will enable constraint-aware, continuously improving process operation.

Ultimately, this approach will reduce variability in fibre side streams, enhance digestibility and functional performance, and offer a scalable solution for industrial biorefineries and feed applications.