The junction of computer programming and chemical engineering has provided ubiquitous tools for engineers to study chemical reactions within microfluidic reactors. However, a future with autonomous laboratory-scale reactors that can recommend requires new advancements. In this study, a MATLAB program was written to track individual droplets within steady-state liquid-liquid segmented flows captured in real-time, frame-by-frame, using a camera. Three different residence times (2.0, 5.0, and 10 min) of a liquid-liquid system (
e.g., silicone oil-water dosed with UV-active (~260 nm) fluorescene) were investigated using a spiral micoreactor with straight sections. Two algorithms were written for analyses of the resultant aqueous phase in recorded videos (at 30 fps). The system was first calibrated according to the 2D geometry and characteristic length scales (
e.g., 300 μm channel width). The first algorithm, the more crude approach, converted each image into a binary (black and white) image and used a connective component analysis (CCA) to split each cluster of pixels into its own structure. This allowed a rigorous frame-by-frame analysis. Distributions of length, velocity, and color were mapped in MATLAB under CCA. Here, the average pixel length along the width of the 300 μm wide channel was found to be 10 pixels, which was axially consistent even throughout the turns. The measured distributions were within reasonable size, confirmed by measurement with a caliper during an experiment. Due to the extensive amount of frames (30 fps), the color distribution mapped was accurate and only limited by the CCD detector resolution. Velocity distributions for each 2.0, 5.0, and 10 min residence times were available
viaa subtraction algorithm that compared adjacent frames of videos of each residence time. The method of CCA also used a reference âcentroidâ, calculated as a weighted average of pixels in a 2D array, that deviated from the actual center of the slug when the slugs traveled through turns.
A second algorithm of analysis along the parameterization of a reactor path was developed. Utilizing each video of the spiral microreactor, any 1D path could be generated. Along the 1D path, color distributions per slug, length, and velocity distributions along the axis of the reactor for each residence time were estimated and found to be more accurate than the first algorithm. The average slug size for each 2.0, 5.0, and 10 min residence times was calculated. The center of the reactor, in the second algorithm, was calculated precisely through a rigorous image-processing algorithm limited solely by the resolution of the camera. This method of skeletonizing a path specified by user parameters calculates the center of the reactor, provided the reactor is continuous and contains no branches. The 1D parameterization of any continuous microreactor geometry using a 5-20 sec video of multiphase flows, provided the residence time is >20 sec, is possible using the algorithm. The same routine was then tested by the Briggs-Rauscher oscillating reaction occurring within the dispersed slugs in order to determine useful kinetics. This non-invasive, visual method of data acquisition could revolutionize the way we discover catalyst and kinetics for certain classes of chemistry.
