Beverage mixing requires reliable monitoring of concentration distribution, mixing state, and product quality during preparation. However, conventional measurement methods often provide only limited point-based information and may not fully describe how mixing evolves over space and time. This paper proposes a knowledge-driven decision-making support framework for optical analysis of beverage mixing. The proposed system processes optical experiment data to extract mixing states and Brix concentrations, then organizes them into a Neo4j knowledge graph. The Neo4j database backs up a graph-grounded chatbot, which then queries this base to deliver user-oriented explanations, recommendations, and decision support. A Planar laser-induced fluorescence (PLIF) based case study using the HFCS solution and Rhodamine dye is presented to demonstrate the workflow from optical data acquisition to machine vision analysis and to back up knowledge-based reasoning. The results show the possibility of combining machine vision and knowledge graph reasoning to improve interpretation, troubleshooting, and future automation of beverage mixing analysis.

Keywords

Optical mixing analysis, Beverage mixing, Knowledge graph, LLM-assisted reasoning, Brix estimation