Robotic system for automated visual detection of physical instability in protein systems

Evaluating the physical instability of protein dispersions is critical in food science and product development, yet current methods rely on subjective visual inspection or costly light-scattering instruments. These approaches provide only snapshots in time and offer limited temporal resolution. We present a robotic and computer vision system for automated assessment of the physical instability of protein dispersions across varying pH conditions. The system automates titration and pipetting tasks, while an optical camera continuously monitors up to ten samples stored in cuvettes. Captured images are analyzed with a Laplacian-based algorithm that enhances agglomerate visibility and quantifies instability through a single metric. The system was validated through acid titration of a 1% milk protein solution at 30 °C, successfully detecting both massive physico-chemical instability (aggregation) and subtle physical instability (sedimentation). The amount of added acid required to reach the instability point showed a 6% variation in the millimolar range, demonstrating high reproducibility across experiments. This cost-effective and scalable platform enables parallel analysis of up to ten samples, making it well suited for early-stage formulation screening. Beyond protein dispersions, the system could be extended to study more complex, multi-ingredient formulations. In addition, it can facilitate the generation of large, high-quality datasets to support the development of machine-learning models for formulation optimization.