Integrating machine learning and drying dynamics of complex fluid droplets for pattern recognition

Droplets of complex fluids drying on solid surfaces generate rich spatiotemporal patterns arising from nonlinear coupling between evaporation, transport, interfacial stresses, and phase transitions. While these patterns have long been studied from a physical perspective, recent advances in machine learning (ML) provide a robustframework for systematically decoding, classifying, and predicting drying behavior across diverse systems. This review highlights how ML learns from drying-droplet images and predicts previously unseen patterns. We review how physics-guided feature extraction, including texture statistics, crack morphology, height profiles, and flow-derived metrics, enables traditional ML methods to link final dried patterns with fluid composition and initial conditions. We then emphasize deep learning approaches that bypass manual feature design, learning hierarchical representations directly from images and time-lapse sequences. Beyond static end-point images, we underscore the emerging role of time-resolved ML in capturing the entire drying dynamics, enabling robust classification and predictive forecasting. Finally, we discuss interpretability, generalization, and reproducibility as critical challenges, and outline future directions toward multimodal, physics-informed, and interpretable ML frameworks. By integrating ML as an analytical tool rather than solely as a post-processing step, this review discusses how ML-based analysis of drying droplets enables scalable, low-volume approaches for pattern-based screening applications.