Project Description

Crystallization not only determines product purity, but also—through crystal size and shape—filtration characteristics, further processing, and, in the case of active pharmaceutical ingredients (APIs), the solid-state form relevant for regulatory purposes. Despite formally identical process control, industrial practice shows significant batch-to-batch variations. The cause is suspected to be hard-to-access interfering factors (impurities, raw material batches, history of upstream process steps, operator and supplier influences) that are not detected by standard sensors.

The goal is to develop an explainable, image-based inline monitoring system for batch cooling crystallization that detects deviations from target quality during the ongoing process and identifies their causes, rather than identifying them only retrospectively in the final product.

Key Elements of the Project

• Validation of the optical imaging-based inline measurement technology on various material systems in a batch stirred-tank crystallizer
• Digital twin as a data source through the development of a first-principles model with a population balance
• Curated reference database consisting of image, process, and contextual data as the basis for training
• Development of an early-warning system based on deep neural networks (time-series and contextual data) that detects deviations in the batch process at an early stage
• Explainability through causal hypotheses for detected deviations instead of a black-box score

Project Description

This project, part of the VIVET innovation community, aims to establish an integrated approach for the development of energy- and resource-efficient chemical syntheses. The goal is to provide methodologies and technologies for evidence-based, model-driven chemical process development. These methodologies will be tested using two model systems – the synthesis of benzimidazoles and naphthohydroquinones – to enable environmentally friendly production routes. Comprehensive knowledge of the underlying physicochemical processes and their reliable representation through mathematical models is crucial for the efficient and sustainable operation of these syntheses. To support this, a differential recycle reactor has been developed at LRF, which enables kinetic studies under well-defined thermal, chemical, and fluid dynamic conditions.

Project Partners

The EnaChem project is carried out in close collaboration with the University of Applied Sciences Mannheim (TH Mannheim). While LRF investigates the reduction pathways of the syntheses, TH Mannheim focuses on the corresponding oxidation.

The industrial partners within the VIVET innovation community are diverse and span various sectors. In addition to partners from the chemical and pharmaceutical industries, such as BASF, Merck, and Roche, companies from the field of measurement and control technology – including IPS and HiTec Zang – are also part of the community. Other organizations and associations, such as the IHK Rhein-Neckar and the Smart Industries network, complete the consortium.

Key Elements of the Project

• Design and construction of a universally applicable screening apparatus for hydrogenation reactions
• Systematic screening of catalysts and solvents
• Development and calibration of a quantitative analytical method
• Execution of kinetic studies in a differential recycle reactor
• Investigation of various hydrogenation syntheses
• Development of a kinetic model
• Model-based scale-up of the reaction
• Knowledge transfer and communication with the industrial partners involved