ISBN 9783843955065

978-3-8439-5506-5, Reihe Apparatedesign

Laura Maria Neuendorf
Artificial Intelligence-based Process Supervision for Analysis and Automation

242 Seiten, Dissertation Technische Universität Dortmund (2024), Softcover, A5

Zusammenfassung / Abstract

The introduction of artificial intelligence (AI) methods into process engineering contributes to the improvement of modeling processes, product properties, and facilities. Continuous, AI-based real-time process monitoring enables the analysis of complex multi-phase interfaces such as droplets and crystalline particles. More information can be obtained in a shorter time, data of higher quality can be measured, and previously immeasurable states can become measurable and thus processes can be automated. This automated process monitoring and control is not only effective but also increases understanding of the operations and helps reduce human errors.

In this study, it was experimentally investigated how the application of AI algorithms for process monitoring and automation can be beneficial. The goal was to develop novel image analysis methods for laboratory measurement procedures and improve the effectiveness of sensors in chemical plants through AI integration. The focus was on three main applications: the analysis of crystalline particles, liquid-liquid phase interfaces, and especially droplets. Various parameters were identified that are of great interest for process analytics.

The examination of liquid-liquid extraction columns was particularly enlightening. Structures such as overlapping droplets and specific operating conditions, such as column flooding, could not be automatically measured and monitored before. Through the integration of AI image analysis methods, it is now possible to measure droplet dimensions in real-time and fully automate the operation of the extraction column.

Such automation is also conceivable for other process engineering operations and challenging-to-adjust phenomena. The developed AI image analysis methods, therefore, have the potential for transferability to a variety of process improvements.